In this article, we will introduce AI agents, which have been gaining popularity since 2025, by explaining (1) what an AI agent is, (2) what Web3 AI agents are, and (3) the legal issues surrounding AI agents.
Because AI agents can take over any job, when considering the relationship between AI agents and the law, it is necessary to consider legal issues for all the jobs they perform. However, since it is difficult to cover all of this in a blog, this article first introduces the basic ideas for considering legal issues related to AI agents, and then discusses their relationship with regulation, focusing on financial regulation in particular.
However, we believe that this approach to financial regulation will also be useful to a certain extent when considering legal issues related to other AI agents.
An AI agent is generally an artificial intelligence system that autonomously performs specific tasks. It is capable of processing data from the environment, learning and making decisions as needed, and carrying out tasks without human instruction.
Generally, an AI agent has the following elements:
(i) Recognition: Processes the external environment and input data to understand the current situation. (ii) Decision-making: Plans actions to accomplish tasks based on data. (iii) Action: Executes actions that bring about changes in the environment based on the plan. (iv) Feedback: Uses the results of execution to learn and improve actions next time. |
This allows AI agents to perform repetitive tasks and make complex decisions on behalf of humans.
AI agents are currently attracting a great deal of attention as they are expected to transform our lives and how we do business.
For example, AI agents are expected to be used in the following applications:
Examples of AI Agent Uses: (i) Generative AI Creativity Support: Used in creative fields, such as generating text, images, videos, and music. In the media, advertising, and gaming industries, AI is expected to improve production efficiency and create new value. (ii) Personal Assistant: AI can provide support tailored to individual needs, such as life coaching, educational support, and business assistants. Applications that improve daily life efficiency, such as schedule management and health advice, are gaining attention. (iii) Autonomous Use in Finance: AI agents that support asset management and household finances use data to propose optimal investment strategies and savings methods. Automated trading and asset management are also advancing in decentralized finance (DeFi). (iv) Business Process Automation: Automating repetitive tasks such as human resources, finance, and customer service contributes to improving corporate productivity. Data analysis and decision-making support are also areas where AI agents excel. (v) Healthcare: AI agents are used for health management, telemedicine, and disease prediction. In particular, personalized services such as symptom analysis and mental health support are expected. (vi) Autonomous Systems: AI agents are expected to play an active role in handling physical tasks such as robot automation in warehouse management, logistics, and disaster response, as well as self-driving and drone operation. |
AI agents have the potential to enrich our lives and make businesses more efficient, thanks to their personalization and autonomy, and these applications are areas where further advancements are expected in the future.
The following are some specific examples of AI agent use both in Japan and overseas.
Service name | Provider | Features |
Fujitsu Kozuchi AI Agent | Fujitsu Limited | An AI agent that autonomously promotes advanced tasks in cooperation with humans. For example, as a conference agent, AI can participate in meetings to share information and propose measures, or as a field support agent, analyze camera footage at manufacturing and logistics sites, propose improvements, and create work reports. |
Agentforce | Salesforce, Inc | Autonomous AI assistants. For example, Service Agent, part of Agentforce, replaces traditional chatbots with autonomous AI, enabling accurate and fluent conversations with customers 24/7 without pre-programming scenarios. |
Operator | OpenAI, Inc | AI operates a web browser on behalf of the user, automating everyday tasks. It uses its own browser to navigate web pages and perform operations such as typing, clicking, and scrolling according to the user’s instructions, allowing it to automate tasks such as making restaurant reservations and online shopping. |
Pactum AI | Pactum AI, Inc. | Walmart has introduced Pactum AI, an autonomous negotiation AI, to automate negotiations with over 100,000 suppliers. It automatically makes proposals in response to requests from suppliers based on pre-specified budgets and priorities, leading to optimal trading terms for both Walmart and the supplier. |
Waymo Foundation Model | Waymo LLC | Waymo, which operates self-driving taxis, uses a proprietary AI model called the Waymo Foundation Model to enable advanced decision-making, from understanding the surrounding situation to generating driving plans. |
AI agents are said to be compatible with Web3. The integration of Web3 and AI agents is expected to create new possibilities, such as the following:
Examples of Web3 AI Agent Use Cases : (i) Integration with Decentralized AI Agents and Smart Contracts: AI agents operate smart contracts on the blockchain and execute transactions autonomously. For example, real estate and financial transactions can be completed without an intermediary. ・Operating as part of a Decentralized Autonomous Organization (DAO): AI agents participate in the decision-making process within the DAO, making proposals and voting. (ii)Strengthening User Sovereignty and Protecting Privacy: AI agents process user data locally and securely store personal information in decentralized storage (e.g., IPFS). ・Self-Sovereign Identity (SSI): AI agents leverage users’ SSI to simplify access and authentication to Web3 services. (iii) Automating the Token Economy and Automating Token Trading: AI agents manage and trade assets on behalf of users on a decentralized exchange (DEX). ・Reward Distribution: AI agents receive and redistribute tokens based on the value they generate on the Web3 platform. (iv) Metaverse and AI Agents: AI agents act as virtual assistants within the metaverse. For example, managing land for users or trading NFTs. (v) Utilizing Zero-Knowledge Proofs (ZKPs): AI agents can use ZKPs to provide trust in Web3 applications while protecting privacy. |
One example of a Web3 AI agent that has attracted global attention is AI16Z (ai16z), a decentralized AI investment fund built on the Solana blockchain that utilizes AI agents to autonomously conduct investment activities.
Project name: ai16z Platform: Solana Blockchain Features: ・AI collects and analyzes market information and automatically executes token transactions taking into account community consensus. ・Uses decentralized governance that allows investors to participate in project management and decision-making through tokens. ・Blockchain technology ensures the transparency and reliability of investment activities. AI Agent “Eliza”: ・An AI agent responsible for planning and executing investment strategies. ・Released as open source, it can also be deployed by third parties. |
The name ai16z is a play on Andreessen Horowitz (a16z), a well-known Silicon Valley VC, but ai16z and a16z are unrelated.However, on October 27, 2024, Marc Andreessen, one of the founders of a16z, posted “GAUNTLET THROWN” on X (formerly Twitter) and mentioned the T-shirt worn by ai16z’s main avatar, which caused the name of ai16z to spread rapidly1.
Furthermore, in early January 2024, ai16z’s market capitalization temporarily exceeded 300 billion yen, growing more than 100 times in three months. As a result, by early January 2025, the company had become a hot topic worldwide and in the case of a personal injury accident was also a major topic of discussion on AI agents on Japan’s X (formerly Twitter).However, perhaps due to overly high expectations, the price has fallen sharply, with the market capitalization falling to around 50 billion yen, indicating extremely speculative price movements.
1 AI Agents and Legal Regulation (Basic Concept) (i) The word “agent” in AI agent translates to “proxy” in Japanese. Even if a service called an AI agent does not strictly qualify as an “agent” in the legal sense, it typically refers to an entity that performs specific tasks “on behalf of” a human. (ii) When considering regulations applicable to AI agents, we first consider (1) what regulations would be imposed if a human were to perform similar actions, (2) whether any regulations would be imposed on the user if the user were to perform such actions using AI, and (3) whether any regulations would be imposed on the business if the business provides the AI to users. (iii) Note that, in cases such as DAOs, if the AI agent can be said to operate completely autonomously and without human involvement, there is room to consider that legal regulations do not apply in the first place. However, since it is often unclear whether there are no operators who would be subject to regulation, careful consideration is required. 2 Relationships between AI Agents and Users, and Relationships between AI Agent Providers and Users (i) When a part of a task is delegated to a human, it can take the form of ① outsourcing (quasi-agency or subcontracting), ② labor dispatch, or ③ employment. However, when a part of a task is delegated to an AI agent, no contractual relationship arises between the AI agent and the user; the human is simply effectively using the AI agent. (ii) The relationship between an AI agent provider and its user is governed by contractual relationships, such as a service agreement (e.g., SaaS) or a system development agreement for the AI agent. 3 AI Agent Errors and Liability (i) Discussion of Holding Providers Liable The relationship between an AI agent provider and its user is governed by contracts and regulations. If an AI agent malfunctions, the service provider providing the AI agent may be held liable for breach of contract or other issues. (ii) Ordering Errors by an AI Agent (Unauthorized Orders or Unauthorized Agency) ①If an AI agent managed by a user places an incorrect order, the effects of the order generally belong to the user. Depending on the instructions given to the AI agent, its behavior, settings, and management status, there is theoretically room for consideration of canceling an order due to error. However, from the perspective of transaction security, such a claim would likely only be accepted in extremely limited cases. ②For AI agents managed by others, the issue is whether the AI agent provider acted without authorization. Regarding apparent agency, for example, if an AI agent is given a password or has order authority and transacts beyond that authority, the other party will have no choice but to consider the transaction legitimate, and apparent agency would essentially be established. ③For example, if an AI agent provided by a financial institution such as a cryptocurrency exchange or securities company malfunctions and places an erroneous order, a user may be able to sue the financial institution for damages or claim cancellation due to error. Therefore, although it would reduce convenience, taking measures such as requiring a human (user) to personally confirm the final order details would be effective in addressing the risk of erroneous orders. (iii) Liability for Damages Caused to Others through the Use of AI Agents For example, if damages are caused to others through the use of an autonomous driving AI agent, who is liable? ① The driver of an autonomous vehicle owned by the user may be liable for damages under the Automobile Liability Act or the Civil Code; ② The automobile manufacturer may be liable for damages under the Product Liability Act (PL Act); and ③ The software provider providing the AI agent may be liable for damages under the Civil Code. 4 Web3 AI Agents and Financial Regulation (i) If an AI agent trades cryptocurrencies or stablecoins on behalf of users on a DEX, it is necessary to consider whether regulations for cryptocurrency exchanges and electronic payment instruments trading businesses apply to the provider of the AI agent. While mere assistance to users is not subject to regulation, if the AI agent is deemed to act as an intermediary, it may become subject to regulation. (ii) Investment advisory and management services for spot trading of cryptocurrencies and stablecoins are currently not subject to regulation under the Financial Instruments and Exchange Act. Therefore, even if an AI agent is conducting such transactions, these regulations are generally not applicable. On the other hand, investment advisory and management services for cryptocurrency and stablecoin “derivative trading” are subject to regulation under the Financial Instruments and Exchange Act. Even when services are provided by AI agents, the provider may be subject to regulations under the Financial Instruments and Exchange Act. (iii) If a GK uses an AI agent to trade spot cryptocurrency or stablecoins when conducting fund management operations, such as in a GK-TK scheme, regulations under the Crypto Asset Exchange Business and Electronic Payment Instruments Business may not apply. On the other hand, if another company receives investment discretion from the GK and uses an AI agent to conduct such operations, regulations under the Crypto Asset Exchange Business and Electronic Payment Instruments Business may apply. 5 Other Laws (i) When an AI agent provides customer service, measures must be taken keeping in mind the warnings regarding AI issued by the Personal Information Protection Commission. In relation to Article 4 of the Consumer Protection Act, measures must be taken to prevent hallucination. |
The “agent” in AI agent is translated as “representative” in Japanese. Services called AI agents usually refer to entities that perform specific tasks “on behalf of” humans, even if they do not fall under the strict legal definition of a “representative.”
When considering regulations that may apply to AI agents, consider the following steps:
(i) Consider what legal issues would arise if a human were to perform a similar act.
(ii) Then, consider whether any regulations would be imposed on users if they were to perform such acts using AI.
(iii) Consider whether any regulations would be imposed on businesses that provide the AI to users.
As mentioned above, AI agents are sometimes translated as “agents,” but since they are neither people nor corporations, under current law, AI agents themselves are not subject to regulation. Instead, the natural persons and corporations that use or provide them are.
In relation to this “natural persons and corporations are subject to regulation,” particularly in the context of DAOs, if an AI agent can be said to operate completely autonomously and without human involvement, the question arises as to whether legal regulations would not apply in the first place. However, since it is often unclear whether there are no operators who are completely subject to regulation, we believe that careful consideration is necessary2.
Currently, there are no laws that generally prohibit the provision or use of AI agents, so whether or not the current regulations that apply to natural persons and legal entities apply to each individual act must be considered.
In relation to point 2 above, even if we translate “agent” as “representative,” AI is neither a natural person nor a legal entity, and AI itself cannot be the subject of rights or obligations.
Therefore, for example, when an AI agent makes a mistake, the AI itself is not liable, but the user or the AI agent provider is.
AI agents are automating a variety of tasks.
First, when a person delegates part of a task to another person, a contract of the following form is concluded.
Relationships between people (i) Outsourcing (quasi-agency/subcontracting) ● Generally suitable when outsourcing short-term work. ● Contracting (Article 632 of the Civil Code) is used when a specific output or task is required to be completed, and quasi-agency (Article 656 of the same Code) is used when a specific task is required to be performed. ● Main relevant laws and regulations: Subcontracting Act, Antimonopoly Act, Freelance Act, etc. (ii) Worker dispatching ● Generally suitable when temporarily supplementing one’s own staff. ● Workers are employed by the dispatching company and perform work at the dispatched company. ● Main relevant laws and regulations: Worker Dispatch Act, etc. (iii) Employment (Article 623 of the same Code) ● Generally suitable when securing a stable workforce for ongoing work. ● Main relevant laws and regulations: Labor-related laws and regulations such as the Labor Standards Act. |
On the other hand, under current law, the relationship between humans (users) and AI agents is merely that between humans and the software and hardware (that construct the AI agent) and is not a contractual relationship; it is merely a relationship in which humans are effectively using the AI agent.
AI agents are generally developed by companies, and many users either receive ready-made AI agents from those companies or commission the development of AI agents to those companies.
This relationship can be summarized as follows:
(i) Obtain a license to use the AI agent provided by a service provider such as a SaaS service provider and use it in compliance with the terms of use. (ii) Develop, install, and operate an AI agent system for your company. |
If a user suffers damage due to a malfunction of an AI agent, the following liability claims and defenses may be taken:
The user’s argument: ・Based on the contents of the SLA (Service Level Agreement), etc., the user may seek compensation for damages (Article 415 of the Civil Code) or terminate the contract (Articles 541 and 542 of the Civil Code). The service provider’s possible arguments: ・There are exemptions and limitations on liability based on the terms of use; ・The service provider is not at fault (Article 415, Paragraph 1, of the same law); ・The user is also negligent (Contributory Negligence, Article 418 of the Civil Code). |
For example, if a person asks another person to purchase Bitcoin and gives them authority to do so, but the agent ends up purchasing Ethereum, this will be considered unauthorized agency, and in principle the effects of the contract will not belong to the principal.
The main legal issues that arise when unauthorized agency occurs are as follows:
● Ratification of unauthorized agency (Articles 113 and 116 of the Civil Code) ● Liability of unauthorized agent for performance or damages (Article 117 of the same Code) ● Application of apparent agency (Article 110 of the same Code) ➡ If the counterparty has “legitimate reasons” to believe that the principal has the authority of agency, the effects of the contract may belong to the principal. For example, if the agent has the means to prove the authority of agency (possession of a registered seal or power of attorney, etc.). However, in the following cases, if the counterparty does not conduct an appropriate investigation or confirmation of the existence of the authority of agency, it may be determined that there is no “legitimate reason” and the apparent agency may not be established. ✓ If there are signs of tampering with the power of attorney ✓ If the seal on the power of attorney is a cheap seal ✓ If the transaction is disadvantageous to the principal |
Since AI agents are programs that operate based on user instructions, orders placed by AI agents are generally considered to be an expression of the user’s intention, and the effects of such orders are also considered to belong to the user.
However, there are also cases where an AI agent places an order that differs from the user’s true intention, and in such cases, the question arises as to whether the effects of the order belong to the user.
In this regard, it may be necessary to consider whether the user can revoke the expression of intention as a “mistake” (Article 95 of the Civil Code). There are two cases of mistake:
1. Mistake resulting in a lack of intention corresponding to the manifestation of intention (Article 1, Paragraph 1, Item 1) In principle, rescission is possible if the mistake concerns an important matter. 2. Mistake resulting in an untrue understanding of the circumstances that formed the basis of the legal act (Article 1, Paragraph 2) In principle, rescission is possible only if the mistake concerns an important matter and the circumstances were disclosed to the other party. |
(a) When the user’s instructions and the order result match
For example, if a user intends to “purchase cryptocurrency at the discretion of an AI agent” and issues such instructions, resulting in the purchase of an unexpected type and quantity of cryptocurrency, since the user’s intention to “purchase cryptocurrency at the discretion of an AI agent” and the result match, it can be said that there is a user intention corresponding to the expression of intention (the AI agent’s order), and unless the other party is informed that “the user thought the AI agent would operate within the user’s expectations,” it is likely to be difficult to cancel the order due to mistake (paragraph 2 of the same article).
(b) When the user’s instructions and the order result do not match
On the other hand, if the user gives specific instructions specifying the type and quantity, and the AI agent places an order for a different type and quantity, it seems theoretically possible to argue that the cancellation was due to mistake, on the grounds that the user’s intention (the AI agent’s order) does not correspond to the expression of intention.
However, if such cancellations were easily permitted, it would likely seriously undermine the safety of transactions. Therefore, Article 95, Paragraph 3 of the Civil Code stipulates that cancellations cannot be made if the user is “grossly negligent.” For example, if there is a setting error or improper management of the AI agent, the user could be found to be “grossly negligent” and the cancellation could be denied. In the case of orders placed by a company, it may even be considered “gross negligence” if the user does not check the specific order contents themselves after using the AI agent.
(c) Special provisions for electronic consumer contracts
When consumers place orders using AI agents, Article 3 of the Electronic Consumer Contract Act (Act on Special Provisions of the Civil Code Concerning Electronic Consumer Contracts) is likely to apply. This Act, because of the high incidence of ordering errors in internet transactions, allows cancellation in the following cases in principle:
① Orders made by mistake (e.g., pressing the “Buy” button by mistake) ② Orders made by incorrect input (e.g., entering the wrong quantity) ③ Orders made differently from your intention due to automatic input or incorrect operation |
Even when using an AI agent, if a consumer conducts a transaction using a computer in accordance with the procedures displayed on a computer screen by a business operator, this will be considered an electronic consumer contract (Article 2, Paragraph 1 of the same Act), and it is considered that this article also applies to transactions using an AI agent.
However, this exception will not apply if a business takes measures to request confirmation of the consumer’s intention, as described below.
① When a final confirmation pop-up asking “Do you want to confirm your purchase?” is displayed. ② When a confirmation screen is set up via the cart instead of one-click purchasing. ③ When the intention to purchase is confirmed using a system such as two-factor authentication. |
Furthermore, if a consumer uses an AI agent to conduct a transaction without taking the necessary confirmation measures, this may fall under the provision of the same article, “where the consumer expresses their intention to the business operator that they do not need to take such measures,” and the special provisions may no longer apply. In such cases, cancellation due to mistake is generally not permitted.
If an AI agent provided by another party is used and the AI agent conducts a transaction that the user did not intend, the issue of unauthorized agency by the AI agent provider may arise. When
the AI agent provider is an unauthorized agent, the following particular issues arise regarding the success or failure of apparent agency.
●In a typical agency relationship, whether the agent has a registered seal or a power of attorney is the key to determining whether there is “reasonable cause” to believe in the existence of agency authority over the other party. ●In the case of AI agents, transactions are digitalized, and it is common for there to be no use of a registered seal or presentation of a power of attorney. |
Therefore, the question arises as to what constitutes a “legitimate reason” for the trading partner. If the AI agent is given, for example, a password or authority to place an order and uses it to place an order, the other party will basically have no choice but to believe that a legitimate transaction has been made, and it would appear that apparent agency would be established.
Column: Cases of erroneous orders made by AI agents provided by financial institutions such as cryptocurrency exchanges and securities companies and available within their services. |
1. When the counterparty is a third party. 2. When the counterparty is the financial institution itself. 3. Implementing User Confirmation as a Risk Avoidance Measure |
If another party suffers damages related to the malfunction of an AI agent, the provider of the AI agent or the user of the AI agent may be liable for damages.This point is particularly noteworthy as an example of the use of AI agents, and autonomous driving is a typical use case in which the use of AI agents could cause harm to others. In autonomous driving, an AI agent will be responsible for driving the car, but legal liability in the event of an accident may differ between when a human is driving and when an AI is driving.
In the event of a personal injury accident, the car owner or other operator (a person who operates a car for themselves) will be held liable under Article 3 of the Automobile Liability Insurance Act (Automobile Liability Insurance Act) in addition to tort (Article 709) of the Civil Code. When claiming damages under Article 3 of the Automobile Liability Insurance Act, the victim does not need to prove the driver’s negligence. Under Article 3 of the Automobile Liability Insurance Act, the operator can be exempt from liability if they meet all of the following three exemption requirements:
(a) The driver and the victim were careful in operating the vehicle. (b) The victim or a third party other than the driver acted intentionally or negligently. (c) The vehicle had no structural defects or functional impairments. |
Since the Automobile Liability Act does not apply to property damage accidents, victims must file a claim for damages based on tort liability in Article 709 of the Civil Code. In this case, the victim must prove the driver’s intent or negligence.
Even if an accident resulting in injury or death occurs due to the autonomous driving of an AI agent, the Automobile Liability Act is generally considered to apply.3 If there is a malfunction in the system of a fully autonomous driving AI agent, the victim may be entitled to claim damages from the operator under the Automobile Liability Act, as it would not meet requirement (c) of the above exemption requirements.
We believe that operators who have paid compensation and insurance companies who have paid insurance claims due to system failures caused by AI agents will seek compensation from automobile manufacturers, AI system software providers, and others.
In the case of property damage accidents, Article 3 of the Automobile Liability Act does not apply, so claims for damages based on tort liability will be made against the driver, etc. However, with fully autonomous driving, there will be no driver error, etc., so it will be difficult to hold the driver responsible for intent or negligence, and it may be difficult to recognize the driver’s liability for damages.
In this case, if there is a problem with the AI agent system, the victim may seek liability from the automobile manufacturer or software developer that provides it, as follows:
Victims may file a claim for damages against the automobile manufacturer under Article 3 of the Product Liability Act (PL Act).
The PL Act imposes strict liability on manufacturers when a defect in a product causes damage to life, body, or property. However, there are also the following issues:
● Software itself is not movable property, so it does not fall under the category of “products” under the Product Liability Act. However, if a vehicle incorporating the software is deemed to have a defect, the automaker may be held liable under the Product Liability Act.4 ● Because AI-based self-driving systems are sophisticated and complex, it may be difficult for victims to prove a “defect” and a “causal relationship.”● Product liability is recognized based on defects that existed at the time of delivery by the manufacturer, etc., so if a defect occurs due to a software update performed remotely after the vehicle is delivered, product liability may not be recognized. |
Victims may seek compensation for damages against software companies that provide AI agents, citing defects in the AI agents. In this case, because software is an intangible object and therefore product liability does not apply, victims may pursue tort liability under Article 709 of the Civil Code.
In this case, the victim will need to prove the software developer’s intent or negligence, so the hurdle for claiming compensation is likely to be higher than in cases where damages are claimed under Article 3 of the Automobile Liability Act or Article 3 of the Product Liability Act.
In this section, we will consider how financial regulations apply to AI agents in Web3, following the ideas in Section III above. Although we will be examining this in the context of Web3, similar ideas also apply to financial AI agents, such as stock investment AI agents.
It is conceivable that AI agents will trade crypto assets and stablecoins on behalf of users on decentralized exchanges (DEXs). Utilizing such a system is expected to bring the following benefits:
● Fast trading through real-time market analysis ● Data-driven decision-making that is not influenced by human emotions |
On the other hand, when conducting such transactions, it is necessary to consider whether there are any regulations regarding crypto asset exchange businesses, etc.
When buying and selling cryptocurrencies and stablecoins (which are linked to the value of legal tender and redeemable at face value), it is necessary to consider the application of regulations regarding crypto asset exchange businesses (Article 2, Paragraph 15 of the Payment Services Act) and electronic payment instruments trading businesses (Article 2, Paragraph 10, Item 2 of the same Act).
Under the law, trading crypto assets as a mere investor does not constitute a “business” and is not subject to regulation. 5 On the other hand,
sales to the general public or acting as an agent for sales to the public are subject to regulation.
Even if an AI agent buys and sells cryptocurrencies or stablecoins on behalf of a user, there are no particular regulations for the user if the AI agent is used for the user’s own investment purposes.
Also, even if there is a company that provides an AI agent to place buy and sell orders, there are likely to be no regulations if the agent simply assists users with the administrative procedures for buying and selling.
On the other hand, AI agents could act as intermediaries, for example, to easily connect users to DEXs.6 and is deemed to be managed and operated by a party other than the user, the provider of the AI agent may be subject to regulations on crypto asset exchange businesses and electronic payment instrument trading businesses (intermediary regulations).
In the Web3 field, AI agents can develop investment strategies and provide investment advice and asset management services related to spot trading of crypto assets and stable coins, as well as derivative trading of crypto assets and stable coins.
In this section, we will explain the main legal issues that must be considered when AI agents provide such investment services, comparing them with when provided by humans.
When providing investment advisory and management services, different legal regulations apply.
Investment advisory services refer to the business of providing advice on investment decisions regarding securities and derivative transactions by entering into a contract (investment advisory contract) for providing investment advice and receiving compensation.
The key points of regulation are as follows:
●Registration as an investment advisory and agency business is required under the Financial Instruments and Exchange Act (Article 2, Paragraph 8, Item 11, Paragraph 3, Item 1, Articles 28 and 29 of the Financial Instruments and Exchange Act). However, advice provided free of charge is not subject to regulation. ●Advice regarding spot trading of crypto assets and stable coins is not subject to regulation. ●Advice regarding derivative trading of crypto assets and stable coins (which fall under electronic payment methods) is subject to regulation. ●It is necessary to be aware of whether the advice is for spot trading or derivative trading. |
Investment management services are primarily considered to be (a) businesses that invest capital contributions from fund holders primarily in securities and derivative transactions (fund management businesses), and (b) businesses that invest and manage securities and derivative transactions after being entrusted with the authority to make investment decisions and manage assets by customers (discretionary investment businesses).
Key points of the regulations are as follows:
●Registration as an investment management business is required (FIEA Article 2, Paragraph 8, Item 12 (b), Article 2, Paragraph 8, Item 15, Article 28, Paragraph 4, and Article 29). Even if the service is provided free of charge, it is subject to regulation if it constitutes a “business.” ●(a) With regard to fund management business, self-offering generally requires registration as a Type II Financial Instruments Business (FIEA Article 2, Paragraph 8, Item 7, and Article 28, Paragraph 2, Item 1). However, there are exceptions, such as Special Business for Qualified Institutional Investors, etc. (FIEA Article 63). ●(b) If customer assets are entrusted to custody through discretionary investment business, registration as a Type I Financial Instruments Business is also required (FIEA Article 28, Paragraphs 5 and 1, Item 5, Article 29, and Article 42-5). ●If the investment target is spot trading of crypto assets or stable coins (in the case of (a) fund management business, if the investment target is “primarily”), it does not constitute an investment management business. On the other hand, if the investment target is derivative trading of crypto assets or stable coins (which constitutes an electronic payment instrument), it is subject to regulation as an investment management business. ●GK-TK Scheme7, all investments made by anonymous partners belong to the assets of the GK (operator) (Article 536, Paragraph 1 of the Commercial Code), and the GK conducts business in its own name. Therefore, (a) if the GK buys and sells spot crypto assets based on its fund management business, it is generally considered to be a transaction for the purpose of self-investment and therefore does not require registration as a crypto asset exchange business.8 In the case where the investment target is a physical stablecoin, it is considered to be a parallel case and would not be considered an electronic payment instruments trading business. ●(b) In the case of a GK-TK scheme, etc., where a GK entrusts investment operations to another company and the other company also buys and sells crypto assets and stablecoins, there is a possibility that it will be subject to regulations for crypto asset exchange businesses and electronic payment instruments trading businesses.9 |
When an AI agent provides investment advisory or management services, the question arises as to whether that business is subject to financial regulations. Generally, it is thought that the application of regulations to those who provide AI agents will be considered.
The key points of regulation are generally the same as when a human does the work, but the following points are particularly important in the case of AI agents.
●Even when holding customer funds for discretionary investment management, there is a possibility that registration as a Type I Financial Instruments Business may not be required if the customer funds are held in a smart contract that is not operated by the AI agent provider. ● After the AI agent is provided, it may not be subject to regulation, especially if the AI agent operates completely autonomously as a DAO, without the developer being involved in its operation, and investment management is automatically executed by smart contract. |
AI agents could potentially act as virtual assistants, assisting with sales and answering inquiries. For example, when selling products or services within the metaverse, avatars equipped with AI agents could be expected to automatically provide customer service.
In this section, we will discuss the main legal issues surrounding AI agents providing customer service, comparing them with traditional human services.
When humans handle customer interactions, they must comply with laws and regulations, for example, from the following perspectives:
(a) Handling of personal information
When acquiring and using personal information when dealing with customers, you must comply with the following rules of the Personal Information Protection Act.
● Specify the purpose of use as clearly as possible (Article 17, Paragraph 1 of the Personal Information Protection Act) ● Do not use personal information beyond the scope of the specified purpose (Article 18, Paragraph 1 of the same Act) ● Notify or publicize the purpose of use to the individual (Article 21, Paragraph 1 of the same Act) |
(b) Consumer protection regulations
When explaining services or providing information to consumers, you must comply with the following regulations based on Article 4 of the Consumer Contract Act.
●Do not give false explanations about important matters. ●Do not provide definitive judgments about uncertain future matters. ●Avoid intentionally or through gross negligence withholding facts that are detrimental to consumers. |
In the event of any of these breaches, the consumer has the right to cancel the contract, so it is important to provide accurate and sufficient information.
(a) Handling of personal information
Even when AI agents deal with customers, they must handle personal information with care.
The Personal Information Protection Commission has issued a warning to OpenAI service providers, including that they must “notify or publicly announce, in Japanese, the purpose of use of personal information of users and other individuals,” and that they must not acquire sensitive personal information without the consent of the individual.10 In addition, the government has issued a warning to businesses that use generative AI to handle personal information, stating that “when a business handling personal information inputs prompts containing personal information into a generative AI service, it must fully confirm that the input is within the scope necessary to achieve the specified purpose of use of the personal information.”11
When handling personal information using AI agents, it is necessary to keep these precautions in mind.
(b)Hallucination by AI agents
From the perspective of complying with Article 4 of the Consumer Contract Act, etc., there is a risk of “hallucination,” where an AI agent provides insufficient information or gives incorrect answers based on insufficient training data or outdated information.
The following measures can be considered to prevent this problem.
● Continuously train the AI agent using the latest and most accurate learning data. ● Implement a feedback function that allows consumers to report misinformation. ● Operators should check the AI agent’s responses as appropriate and make corrections as necessary. |
Reservations:
Japan has enacted and improved crypto regulations since 2017. Japan was once one of the most crypto-friendly nations in the world, but after 2018, it adopted a stricter regulatory stance. It is, however, now becoming more friendly to the Web3 industry again, with an intention to attract foreign investment.
This article provides an overview of cryptoasset regulations in Japan in 2024.
History of Cryptoasset Regulations in Japan
Early Friendly Era | |
February 2014 | MtGox, located in Shibuya, Tokyo, and the largest exchange in the world, went bankrupt. |
March 2014 | Japanese LDP (Liberal Democratic Party, a governing party in Japan) discussed with the government and decided not to regulate virtual currency at that stage but asked the industry to form a self-regulatory organization. |
May 2016 | Japan enacted the first virtual currency act in the world. The act was made as an amendment to the Payment Service Act (“PSA”). The act was friendly to startups and intended to foster the industry. |
April 2017 | The amended PSA stated above was enforced. |
2017 | There were ICO booms all over the world, and the price of crypto went up. The trading volume of Japanese exchanges became number 1 in the world. Many foreign players came to Japan to start their business. |
Era of Stricter Regulation | |
February 2018 | A massive hacking incident, under which approximately JPY 58 billion equivalent NEM was hacked, happened in Japan (Coincheck incident). |
2018-2021 | After the Coincheck incident, the Japanese government tightened the operation of the regulation. Many exchanges received business improvement orders and suspension orders, and the market became shrunk. |
May 2020 | The amended PSA and the amended FIEA were enforced. |
Era that Web3 became a national strategy | |
2021- | The Japanese government’s national growth strategy in 2021 includes the statement that Web3 became one of the national strategies. Under this strategy, the LDP’s Web3 project team has issued policy recommendations titled the Web3 White Paper8in order to foster Web3 every year since 2022. |
June 2022 | Japan enacted one of the world’s earliest stablecoin regulations. The act was made as an amendment of the PSA and the Banking Act. |
2022 | In 2002, there were collapses of Tera Luna, Three Arrows Capital, and the FTX Group. As a result, the global regulatory environment became stricter. However, Japan had already implemented stringent regulations, which proved effective. (*1) Therefore, Japan did not need to change its regulations even after these collapses.
(*1) Even in the FTX Group’s bankruptcy, the assets of FTX Japan’s customers were all preserved because the regulations required 100% of users’ assets to be segregated. |
June 2023 | Stablecoin regulation was enforced. |
May 2024 | DMM Bitcoin was hacked, resulting in a loss of approximately JPY 48.2 billion worth of Bitcoin. However, we have not seen any regulatory tightening in response to this incident at this stage. |
The PSA defines cryptoassets as property value with the following elements:
(i) which is recorded by electronic means and can be transferred by using an electronic data processing system, (ii) which can be used in relation to unspecified persons for the purpose of paying consideration for the purchase or leasing of goods, etc. or the receipt of provision of services and can also be purchased from and sold to unspecified persons acting as counterparties, and (iii) excluding the Japanese currency, foreign currencies, currency-denominated assets, and Electronic Payment Instruments. |
Under the PSA, the Cryptoasset Exchange Service means any of the following acts carried out in the course of trade:
(i) sale and purchase of cryptoassets (i.e., exchange between cryptoassets and fiat currency) or exchange of cryptoassets into other cryptoassets; (ii) intermediary, brokerage, or agency service for the acts described above (i); (iii) management (custody) of fiat currency on behalf of the users/recipients in relation to the acts described above in (i) and (ii) and (iv) management (custody) of cryptoassets on behalf of the users/recipients. |
Sales and purchases of cryptoassets to Japanese residents are not subject to the regulation unless they are conducted “in the course of trade (gyo to shite)”. An act in the course of trade is generally understood to be a repetitive and continuous act vis-à-vis the public. For example, trading in cryptoassets for one’s own investment purposes or taking custody of cryptoassets of a wholly owned subsidiary are not considered acts in the course of trade.
It should be noted that just because your clients are only institutional investors is not considered as it is not in the course of trade.
Whether or not a CESP solicits Japanese residents is also considered an important factor in determining the regulation’s application. The determination of whether solicitation towards residents of Japan is being conducted is made on a case-by-case basis. For instance, actions such as not blocking access to a website from Japan, providing information in Japanese, or introducing products at events in Japan could be considered factors that indicate solicitation towards residents of Japan.
The custodian of cryptoassets shall take the CESP license. According to the FSA guidelines, whether each service constitutes the management of cryptoassets should be determined based on its actual circumstances. Generally, if a service provider can technically transfer its users’ cryptoassets, it falls under the category of the management of cryptoassets. If a service provider does not possess any of the private keys necessary to transfer its users’ cryptoassets, the service provider is basically not considered to manage cryptoassets.
Accordingly, wallet services, such as non-custodial wallets, where the users manage the private key on their own, are not considered to constitute the management of cryptoassets.
An intermediary generally means a factual act that involves efforts to conclude a legal act between two others. Brokerage or agency service means to perform a legal act in one’s own name and for the account or on behalf of another person.
With respect to a purchase and sale agreement of cryptoassets between third parties, the acts of (i) soliciting the signing of the agreement, (ii)explaining the product for the purpose of solicitation, and (iii) negotiating the terms and conditions fall, in principle, under the category of an intermediary.
The mere distribution of product information papers, etc., may not fall under the category of an intermediary and should be considered on a case-by-case basis.
The PSA requires minimum capital, financial requirements, a physical office, a sufficient number of personnel on staff, segregation of assets, an annual audit, a customer identity verification system, accountability to users, protection of person/s’ information, including sensitive information, and, if outsourced, must retain authority. The service provider must be equipped with the systems for adequate operation and legal compliance deemed necessary to operate a Cryptoasset Exchange Service appropriately and securely. Although the applicant must have a minimum capital base of at least JPY 10 million, and it must not be in negative assets, from our experience, the cost of obtaining the license and starting the internet exchange business can be more than JPY 1 billion.
We are often asked by companies interested in entering the Japanese crypto market whether they can start their business by acquiring an already licensed CESP rather than obtaining a new license. The answer is Yes. Regulatory speaking, change of major shareholders is done just ex-post notification and you can start your business after purchasing the already licensed CESP.
The major issue here is that the purchased CESP shall satisfy the governance and compliance levels, which are similar to those a new licensed exchange shall achieve. If one purchases a cheap CESP, which is just having a license but has not done a business actively, to reach these levels might be difficult and time-consuming. Furthermore, if you wish to change the business model or system of the purchased CESP, you must provide an explanation to and obtain approval from the FSA. The cost of purchasing the licensed CESP, combined with this additional expense, can sometimes be comparable to the cost of obtaining a new license. Therefore, careful consideration is necessary.
The PSA requires the users’ cryptoassets to be segregated from the CESP. Further, the CESP shall keep (i) at least 95% of the users’ cryptoassets in cold wallets and (ii) equivalent to 100% minus those kept in the left column of its own cryptoassets in cold wallets. Thus, as a consequence, the CESP shall hold the equivalent of 100% of users’ cryptoassets in cold wallets.
With respect to fiat currency, the CESP shall deposit its users’ fiat currency in a bank account under a different name from where the CESP deposits its own funds.
A CESP must undergo an annual audit of its financial statements and segregation of assets.
Anti Money Laundering law requires CESPs to conduct a know-your-customer of users. Stricter regulations for anti-money laundering came into effect on June 1, 2023. According to the new Travel Rules, when assets over a certain amount are sent by a user, the receiving and sending CESPs must share information about the users. The lack of interoperability in such information-sharing systems has prevented users from sending and receiving cryptoassets between some CESPs.
The regulations applicable to decentralized exchanges (DEX) are not clear. There is an argument that the regulations do not apply to exchanges that are completely decentralized and have no administrator at all, as there is no entity subject to crypto regulations. However, it is necessary to carefully consider whether there is truly no administrator. Further, entities that provide access software to a DEX may be subject to the regulations for being intermediaries.
As stated later in section III. 1, the sale of cryptoassets issued by oneself is subject to crypto regulations. Providing liquidity to a DEX for cryptoassets issued by oneself may also be considered as engaging in the sales of the cryptoassets.
Pure NFTs, such as trading cards and in-game items recorded on blockchains that do not function as payment instruments, are not considered cryptoassets. The FSA states that the distinction between cryptoassets and pure NFTs is as follows:
(i) the issuer of the NFTs prohibits its use as a payment instrument by technical feature or by agreement
(ii) the quantity and price of the NFTs are not suitable as a payment instrument (specifically, one NFT costs more than ¥1,000 or the total number of the NFTs issued is less than 1 million).
Generally speaking, pure NFTs are not regulated in Japan. Please, however, note that whether NFTs are considered as “pure” NFTs needs careful discussion. For example, if an NFT gives some dividend or economic benefit, it might be considered as a security. Further, an NFT, which is linked to real-world assets, might require a discussion of whether regulation of real assets may apply.
Japan was one of the first countries in the world to establish stablecoin regulations. Stablecoins pegged to fiat currency are defined as electronic payment instruments and require a license different from CESP to offer the related service.
Other stablecoins that adjust their value through algorithms could be regulated as cryptoassets or securities. Stablecoins classified as cryptoassets are subject to crypto regulations, while stablecoins classified as securities are subject to securities regulations (FIEA).
ICO (Initial Coin Offering) is an act of issuing and selling tokens to raise fiat currency or crypto assets from the public. ICO is regulated in Japan. The applicable regulations depend on the legal nature of the issued tokens. If the tokens are considered securities, the token issuance will be regulated by the FIEA. If the tokens are considered cryptoassets, the token issuance will be regulated by the PSA.
The issuance of new cryptoasset-type tokens in Japan is generally done by IEO (Initial Exchange Offering). IEO is an act of raising fiat currency or cryptoassets by an entity entrusting the sales of tokens to a licensed CESP. In the case of IEO, if the issuer itself does not conduct sales activity, the issuer does not need to take the crypto exchange license. If, however, the issuer itself wants to conduct sales activity for its new tokens, it needs to have a crypto exchange license, which requires significant cost and time compared to IEO. Several IEO projects have already been launched in Japan.
The IEO process requires examinations by the exchange, JVCEA, a Japanese self-regulatory organization, and the FSA. The examination checks the feasibility of the project for which the funds will be used, the financial soundness of the issuer, and other factors.
SAFT (Simple Agreement for Future Tokens) is a way to raise funds in exchange for the right to purchase tokens to be issued in the future. SFAT targeting Japanese residents is considered to be subject to fund regulation or crypto regulation, depending on the legal nature of the agreement. However, both regulations do not apply unless the act is done in the course of trade, so we may argue that entering into a SAFT with limited numbers of specific persons, such as business partners who will contribute to developing projects, should not be regulated.
SAFE (Simple Agreement for Future Equity) with token warrant is subject to general equity investment regulations, depending on the attributes of involved entities and investors.
Japanese entities sometimes use J-KISS, a Japanese convertible equity, with a side letter that provides tokens.
Generally speaking, we believe staking service for POS tokens is not regulated in Japan. For example, staking one’s own cryptoassets or becoming a validator for ETH is not regulated in Japan.
Not all staking services, however, are exempted from the regulation. If service providers manage the private keys of users’ cryptoassets (we understand some exchanges provide those services), custody regulation may apply. In addition to managing private keys, if the service providers distribute rewards as well as slashing penalties to the users, fund regulations might apply.
We understand that there are some NFT projects that say that they sell NFTs for crypto, and purchasers can stake NFTs, and can get rewards. We understand fund regulation might apply to such cases, especially in cases where staking does not have any actual usage for providing security.
In crypto lending services, a service provider borrows cryptoassets from users for a certain period of time and pays a lending fee in exchange. No regulation applies to that lending because the Money Lending Business Act regulates money lending, but it does not deem cryptoassets as money.12
It should be noted that crypto custody regulations may apply in cases where the service is considered as custody, not lending, even if a service is titled as crypto lending. One factor that distinguishes lending and custody is whether users can withdraw their assets at any time (deposit) or whether there is a specific required time of non-withdrawal (lending).
Mining cryptoassets requires large amounts of electricity. Thus, mining appears to be regulated in some countries, such as Kazakhstan13Regulation of mining in certain areas in Russia is also being discussed14. In Japan, mining itself is not regulated.
A business that collects money from the public to conduct mining operations and then distributes the proceeds from mining to the customers may be regulated under the FIEA as a fund.
Schemes that one sell mining machines, accept deposits of the machines, and promise to pay fees for the mining results may also be regulated under the Act on Deposit Transactions. If the Act on Deposit Transactions is applied, the business must obtain confirmation from the Prime Minister, but it is said that to get such confirmation is nearly impossible. Creating a scheme to avoid such regulation is important.
principle, classified as miscellaneous income. Miscellaneous income is income that is neither interest income, dividend income, real estate income, business income, employment income, retirement income, forestry income, transfer income, or temporary income. The tax rate for miscellaneous income ranges from 5% to 45%, depending on the amount of total income. The maximum tax rate is about 55% when we calculate income tax as well as residential tax and special reconstruction income tax.
Profit generated by cryptoassets transactions is subject to corporate tax which is about 30% depending on the amount of income and how big a company is.
Cryptoassets for which there is an active market must be valued using the mark-to-market method at the end of the fiscal year and are taxable even if companies do not sell them.
This unrealized gain tax treatment became a huge issue in Japan, and many Web3 companies left Japan.
In 2022, the Japanese government decided to reform this unrealized gain tax, and now the tax is not levied if an issuing company of tokens continues to hold its tokens with certain technical transfer restrictions.
In 2023, another tax reform was proposed and approved by the government. Under the reform, an unrealized gain tax is not applied if a company holds tokens with a certain transfer restriction, even in the case that tokens are issued by other entities (including Bitcoin and Ether etc.)
Disclaimer
The content of this article has not been verified by the relevant authorities or organizations mentioned herein and represents only a reasonable interpretation of their statements. Our interpretation of laws and regulations reflects our current understanding and may change in the future. This article is not intended to be legal advice and provides a summary for discussion purposes only. If you need legal advice on a specific topic, please feel free to contact us.
EOD
A DAO is a decentralized autonomous organization that drives a business or project forward using smart contracts without a specific owner or manager. Overseas clients sometimes ask our firm whether they can sell DAO tokens in Japan.
(1) You need to consider Japanese regulations when you sell DAO tokens to Japanese residents, even if you reside outside Japan. (2) Regulations on DAO tokens differ depending on whether they are investment DAO tokens or community DAO tokens. (3) Investment DAO tokens are generally considered “security.” Their sale is usually regulated by the Financial Instrument and Exchange Act (FIEA). The seller must obtain FIEA registration or delegate the sale’s activities to a licensed FIEA company. Some exemptions exist, but they are not easy to use. (4) Sales of community DAO tokens are either (i) unregulated, (ii) regulated by the Crypto Asset Exchange Business Law, or (iii) regulated by the FIEA, depending on the nature of the tokens. |
Below is a chart you need to consider before selling DAO tokens to Japanese residents.
<Chart to be considered>
Community DAOs often issue governance tokens.
To consider the financial regulation of sales of DAO tokens in Japan, we must look at (a) whether the DAO provides some kind of dividend or more than 100% redemption of the principal (“Dividend, Etc.”), (b) whether the DAO has any legal entity nature that is similar to a joint stock company or LLC and whether the tokens represent nature similar to shareholders rights, and (c) whether DAO tokens can be used as a payment method.
Tokens in most community DAO do not have any dividend feature or profit distribution feature for token holders. If there are such features, the regulation on investment DAO tokens will be applied. Please see item 4 below.
If a DAO is structured in the form of a company, which happens rarely, and the DAO token awards member rights of the company to token holders, the right might be deemed as securities. Type 1 Financial Instruments Business Registration is required for the sale of those tokens.
In general, many DAOs are formed without clarification of the legal form. Some DAOs just use a smart contract and do not have any form of legal entity. Under Japanese law, such DAOs may be classified as partnerships or “associations without a juridical person.” The rights of partnerships and associations without juridical persons do not fall under securities unless there are Dividend, Etc.
The sale of DAO tokens without Dividend, Etc. and so on issued by an organization other than a company should be classified as a crypto asset or NFT.
If the tokens fall under the definition of crypto asset, their sale shall be made by a licensed crypto asset exchange business operator.
However, if the DAO token is considered an NFT, there are no restrictions on its sale.
In 2023, the Financial Services Agency (FSA) issued guidelines stating the distinction between a crypto asset (FT) and an NFT: (https://www.fsa.go.jp/news/r4/sonota/20221216-2/20221216-2.html).
The guideline states that if the asset does not have “means of payment” characteristics, it is not a crypto asset and that having “means of payment” characteristics can be determined by the following criteria: In general, if (a) DAO tokens cannot be used as a payment method, and (b-1) the price of the token is more than JPY 1,000 or (b-2) the issued number of tokens is less than 1M, the tokens are considered NFTs, and their sale is not regulated.
FSA’s revised guidelines a) The issuer must make it clear that the tokens are not intended to be used to pay for goods or services to an unspecified person. For, example the terms and conditions provided by the issuer or business operator handling the product clearly prohibit the use of the product as a payment method, or the system is designed to prevent the use of the product as a payment method. b) Taking into consideration the price and quantity of the relevant property value, technical characteristics and specifications, and other factors as a whole, the tokens that can be used to reimburse an unspecified person for the price of the goods and so on must be limited. For example, the tokens must have one of the following characteristics: ・The price per minimum transaction unit must be high enough not to be used as an ordinary means of settlement (JPY 1,000 or more). ・The issued number of tokens divided by the minimum trading unit (issued volume considered after divisibility) must be limited (1 million or less). |
Under the FIEA, DAO tokens that pay Dividend, Etc. generally fall under the category of “electronically recorded transfer rights.” Thus, in order to sell electronically recorded transfer rights, you must either register as a Type 1 Financial Instruments Exchange Business Operator (Type 1 FIEBO) and conduct the sales yourself or have a Type 1 FIEBO do it for you. Furthermore, when soliciting 50 or more persons and issuing 100 million yen or more, the solicitation of the tokens becomes a public offering (Article 2, Paragraph 3 of the FIEA), which requires the submission of complex offering documents and continuous disclosure.
As an exception, if sales are made only to qualified institutional investors (QIIs, professional investors) or wealthy individuals (with financial assets of JPY 100 million or more) who are 49 or younger, and if technical restrictions are in place to prevent other individuals from becoming DAO token holders through resale, then the sale falls under the exception category called “special business for qualified institutions, etc.,” and self-offering only requires a simple notification under Article 63 of FIEA.
We have heard that many investment DAOs in the US are issued to QIIs, and many investment DAOs limit their sales to avoid US security regulations. The problem in Japan is that the definition of “professional investors” is much narrower than in the US. In Japan, one criterion for becoming QII is that corporations and individuals must have at least JPY 1 billion in securities. In the US, a person who has more than USD200,000 annual income can become a QII, and in the EU, a person who has more than EUR500,000 in financial assets (in addition to satisfying other requirements) can be a QII. These requirements are much easier to satisfy than the requirements in Japan.
Japan also has a narrower exception to submit offering documents. In Japan, offering documents are required for offerings exceeding JPY 100 million. In contrast, exemptions apply for offerings less than USD6 million in the US (sales by Reg A Tier 1) and less than EUR5 million in some countries in the EU.
In light of the above, the issuance of investment DAOs is not popular in Japan but is still possible if you obey Japanese regulations.
EOD
A blockchain game is a game that uses the blockchain and uses crypto assets, tokens or NFTs (Non-Fungible Tokens).
In a typical game the following occurs:
(1) user purchases game assets that belongs to the game operator rather than the user,
(2) such game assets cannot be freely transferred, sold, or lent out, and
(3) even time-consuming data disappears after game distribution ends.
Whereas in blockchain games, it is said that the following occurs:
(1) the user is the holder of the token (game asset),
(2) the token can be transferred, sold, or lent out to third parties,
(3) third parties can also use the token, and
(4) as long as the blockchain exists,15the recorded digital assets will exist in perpetuity.
When providing a blockchain game to Japanese residents, the laws listed below should be taken into account. Here is a summary of the laws that pertains to this matter:
(1) Fund Settlement Law and Security Act |
To issue and sell NFT itself is not regulated in Japan. Exception to it is (i) if tokens are deemed as crypto assets, such as fungible token which might be used as a payment method, they might be regulated under the Fund Settlement Law and (ii) if tokens are deemed as security, such as tokens including dividend feature, they might be regulated by the Financial Instrument Exchange Act. |
(2) Act against Unjustifiable Premiums and Misleading Representations (hereinafter refer to Premiums Law or Premium and Representation Law) |
Blockchain Game players might be given tokens, digital currencies, NFTs, digital assets or other gifts which have financial value when a user register, login, play a blockchain game. These gifts might be considered “premiums” or “free gifts” under the Premiums Regulations of the Premiums Law and the value of them are limited. |
If the Play to Earn games allow players to (a) purchase NFTs and (b) earn some reward (e.g., NFTs or tokens) by playing the game, the reward portion may be subject to the prize regulation. There is an argument that depending on the game design, the reward may not be considered an extra (premium) and may not be subject to the Premiums Law. |
The following is a discussion of each legal issue.
(a) Where is the problem?
Under blockchain games, the game operator frequently sells game characters, items, weapons, and land etc. as NFTs to users in exchange for ETH or other crypto assets.
Japanese law does not regulate sales of pure NFT, but regulate sales of crypto assets and sales of securities. Thus, whether sold tokens are not deemed as crypto assets or securities are crucial issue.
(b) What are crypto assets?
Under the Fund Settlement Act, the seller, Crypto Asset is defined as follows:
Definition of Crypto Assets (Article 2 Section 5 of the Fund Settlement Act |
Definition of Type 1 Crypto Asset |
A property value that is recorded in electronic record and transferred electrically, that can be used to pay for the purchase of goods or receive services to an unspecified person, and that can be purchased and sold to an unspecified person (excluding some stable coins and securities). |
Definition of Type II Crypto Assets |
Crypto Asset that is recorded in electronic record and transferred electrically which can be mutually exchanged with the Type I Crypto Asset with an unspecified party (excluding some stable coins and securities). |
Selling or providing custody of crypto assets are highly regulated and must generally be handled by registered crypto asset exchange operator and also it is not feasible for a blockchain game operator to obtain registration.
At this moment, whether tokens are considered as crypto assets is determined by the number of issued tokens and whether the tokens can be used as some form of payment. NFTs are generally not considered as crypto assets, but if a game operator says it is an NFT, and the NFT has payment features etc., it may be considered a crypto asset.
(c) What are Securities?
Japanese Financial Instrument and Exchange Act (FIEA or Security Act) governs the issuance or sales of securities. Securities includes stocks, bonds, mutual funds, collective investment scheme etc. We have been often asked by blockchain game operators whether it is legal in Japan to sell NFTs, such as land, which generate “income” or “dividend”. If NFTs generate income without the participation of players, they may be classified as securities. Thus, when selling profit-generating tokens, the game should consider including features of players’ effort, such as editing land to attract customers.
(a) General Remarks
The crime of gambling under the Penal Code is established by (1) contesting the gain or loss of property profits by (2) winning or losing by chance. In addition, not only money but also “property interest” is considered to be the object of gambling, and rice, land, and debt collection are all considered to be “property interest” subject to the crime of gambling. Crypto assets are also considered to fall under the category of property interest.
Article 185 (Gambling) |
A person who engages in gambling shall be punished with a fine of not more than 500,000 yen or a fine. However, this shall not apply when the betting is limited to betting on objects provided for temporary entertainment. |
(b) Gacha (Loot Box), Reveal and Gambling Law
Some blockchain games have features of Gacha (loot box) and reveal. Users pay money or crypto assets to a gaming company and get NFTs randomly. For example, users pay 1ETH to get a game character NFT. Game characters may include Julius Caesar, Guanyu, Genghis Khan, Napoleon, George Washington etc, and those characters have different strength, powers, rarity etc, and what users can get is not revealed to users.
It was believed that these sales might be considered as gambling because (i) users pay property value, (ii) users receive property value which differs by chance, and (iii) there is winning or losing (of property value) by chance. However, in 2022, blockchain industries talked with a famous criminal law professor and some regulatory authorities and issued the guideline which states certain Gacha and reveal is not considered as gambling. Although the guidelines have no effect to police or criminal courts, the industry is now considering how following the guidelines may reduce the likelihood of criminal penalties.
The requirement is that the issuer and operator of games do not sell the same NFTs at different prices (for example, if the issuer sells NFTs that include Napoleon with 1ETH via Gacha, the issuer is not allowed to sell Napoleon NFT at a different price in another method), and do not buy back NFTs in a secondary market (for example, the issuer cannot buy back Napoleon NFT in 0.5ETH or 1.5ETH). In such cases, there is either winning or losing), and the issuer and operator shall not overstate the value of some NFTs in Gacha over other NFTs in Gacha.
(c) Synthesis
The same theory that applies to Gacha may apply to synthesis, but because synthesis is not discussed in the guidelines, we take a more cautious approach to synthesis.
(a)Initial Start
Developers often ask us of blockchain games if it is possible to give NFTs, game currency, crypto assets, other property to users free as a login bonuses, playing bonus and ranking bonuses etc. When conducting such distribution, it is necessary to consider the relationship with the Premiums and Representation Law.
(b) About the Premiums and Representation Law
The Premiums and Representation Law prohibits the offering of excessive premiums.
Premiums are (1) offered as a means of inducing customers, (2) offered incidental to a transaction, and (3) economic benefits such as goods or money. As the definition of economic benefits is broad, crypto assets, NFTs, in-game currencies, and other benefits might be considered as economic benefits.
Excessiveness will vary depending on whether the sweepstakes is general sweepstakes, joint sweepstakes, or all-inclusive sweepstakes, Still, it will be based on the following criteria to the extent that it is considered relevant to the game.
Description | Example | Limits on Premium and Prizes | |
Total Prizes | Offering prizes to anyone who uses the products or services or visits the store, not through sweepstakes. | Gifts for all purchases, gifts for all visitors, etc. | Transaction value less than 1,000 yen – Premiums up to 200 yen. Transaction value is over 1,000 yen – Premiums are capped at 2/10ths of the transaction value |
General Sweepstakes |
Offering prizes to users of goods or services by chance, such as lotteries, or by the superiority of specific actions. | In-store raffles, quiz and game competitions. | Transaction value less than 5,000 yen – 20 times the transaction value. Transaction value of more than 5,000 yen – 100,000 yen. (Both are capped at 2% of the total expected sales amount) |
(c) Ranking Rewards and the Premiums and Representation Law
In traditional smart phone games, the top-ranking players frequently receive in-game currency. The economic value of in-game currency has been treated as zero or very low by game operating companies, and there is no issue under the Premiums and Representation Law.
In blockchain games, crypto assets and NFTs, which can be sold outside of the game might be given as prizes. In this case, the general sweepstakes restrictions apply. The transaction value determines the amount of the prize. Although it is difficult to determine how much the transaction value is, a reasonable approach would be to set the minimum charge as the transaction value and allow rewards of up to 20 times the minimum amount or 100,000 yen, whichever is lower.
(d) Play to Earn and the Premiums and Representation Law
If we consider a Play to Earn game as a game where players (a) purchase NFTs or game currency and (b) earn some reward (e.g., NFTs or game currency) by playing, the reward portion may be subject to the Premiums Regulation.
However, whether earned NFTs or game currencies will be considered “premium” is unknown. There is an argument that the Premiums Law does not apply to Play to Earn games because the rewards are not “extras (premiums),” but rather the purpose of purchasing NFTs and playing the game itself. Lottery winnings and game-play prizes, for example, are not considered “premiums,” but rather “the purpose” of the transaction itself (gambling law shall be discussed lottery winnings and game-play prizes). This issue has not been resolved in Japan, and careful deliberation is required.
Reserved Matters
The contents of this document have not been verified by the relevant authorities and are merely a description of arguments considered reasonable under the law. It is only the current thinking of our firm, and our firm’s thinking is subject to change.
This document does not recommend using blockchain games or purchasing NFTs.
This document is intended for blogging purposes only. Please consult your lawyer if you need legal advice on a specific case.
Introduction
Clients often ask us whether it is possible to structure a Decentralized Autonomous Organization (DAO) in Japan. Currently, Japan does not have regulations targeting DAOs, unlike Wyoming State or the Marshall Islands. So, we have written this article summarizing what is typically considered when forming a DAO in Japan.
1.1 What is DAO?
A decentralized autonomous organization (DAO) is a new legal structure with no central authority and members committed to acting in the organization’s best interests. DAOs are used to make decisions in a bottoms-up management style and have gained popularity among cryptocurrency enthusiasts and blockchain technology.
1.2 Classification of DAOs
There are several classifications of DAO described below:
1. Investment DAO Investment DAOs are for-profit DAOs aim at co-investing in a project. They are more likely to attract funds than Grant DAOs because they aim to generate profits mainly through “economic capital.” Examples: Genesis DAO, The LAO, BitDAO, etc. 2. Grant DAO The community contributes monies to the grant pool and votes on funds allocation and distribution decisions in a Grant DAO. Innovative DeFi projects are funded using these DAOs, showing how decentralized communities are more flexible with funding than traditional organizations. Examples: MolochDAO, MetaCarteDAO, Aave Protocol, Uniswap Grants, etc. 3. Protocol DAO When tokens serve as a voting metric for implementing any changes in the protocol, such a governance structure represents protocol DAOs. For instance, MakerDAO has revolutionized the DeFi space with its DAI stablecoin. Examples: Maker, Compound, Uniswap, Aave, Yearn, Sushi, etc. 4. Service DAO A Service DAO is a decentralized working group. They can receive tokens as compensation for their projects. Examples: RAID GUILD, DXdao, PartyDAO, etc… 5. Social DAO A Social DAO offers digital democracy where opinions are heard, and people can share common interests. Example: Bored Apes (BAYC) 6. Collector DAO Artists who use nonfungible tokens (NFTs) to create art rely upon collector DAOs to establish ownership of their art. Example: PleasrDAO 7. Media DAO Media DAOs allow product owners of content (i.e., readers) to contribute directly without involving advertisers for the native token as a reward in return for their contributions. Example: Fore Front (FF), Bankless DAO, etc… Source https://cointelegraph.com/daos-for-beginners/types-of-daos |
1.3. Example of an Existing Overseas Law
A few places where DAOs are regulated are Wyoming State and the Marshall Islands. Below is a short description of forming a DAO in the Marshall Islands.
The Legal Form of a DAO on the Marshall IslandsMarshall Islands proposes a non-profit corporation (limited liability company) as a legal entity form for DAO, which stands out from the general practice to establish DAO as a foundation. Such a company is established in compliance with the general corporate law of the Marshall Islands with specific features that:
How does this work?Generally speaking, such a company works as a limited liability company managed by its members. It has three essential constitutional documents: Certificate of Incorporation, Operating Agreement, and Charter of the Company.The Operating Agreement should include the most crucial matters of your DAO management:
You can amend any of these matters by the members’ decision in compliance with the procedure prescribed in the previous version of the Operating Agreement. Registering a Marshall Islands LLC for DAOHere’s what the process of establishing a legal wrapper for DAO on the Marshall Islands looks like:
The above is a reference from Taras Zharan Web 3 Virtual Legal Officer. https://legalnodes.com/article/marshall-islands-llc-as-a-dao-legal-wrapper |
2.1. Points to Consider
When structuring a DAO, one must consider the financial regulations and the legal form characteristics.
Here are several points to keep in mind:
1. Security regulation under the Financial Instrument and Exchange Act (“FIEA”) may apply when tokens have the possibility of dividends or redemption of the principal of more than 100% (dividends and redemption of the principal of more than 100% are from now on collectively referred to as “dividends, etc.”). As a general rule, token sales of such DAO must be conducted by a Type 1 Financial Instrument and Exchange Business Operator (“Type I license”) or by obtaining a Type 2 Financial Instrument Exchange Business Operator license (“Type II license”).
2. When selling Fungible Tokens without dividends, etc., it is necessary to have a Crypto Asset Exchange Operator conduct the sale or to obtain a Crypto Asset Exchange Operator license.
In contrast, these financial regulations generally do not apply when selling NFTs without dividends, etc.
3. You also need to consider the tax benefits. If you want to pursue tax advantages in an Investment DAO with dividends, etc., you can use a partnership or GK-TK scheme. If tax advantages are not particularly important, an association without rights, a general incorporated association, or a limited liability company can be considered a scheme to issue tokens. For the issuance of Fungible tokens or NFTs without dividends, etc., it may be better to have no particular legal structure.
2.2. Reference Table of Conclusions
The table below summarizes the legal scheme and financial regulations that should be considered in establishing a DAO.
The following regulations apply to token sales of Investment DAOs with dividends, etc. (assuming dividends or principal redemption of more than 100%).
Type of Member’s Rights | Form under Japanese Law | Free distribution of Tokens | Token Sale | Investment Management |
DAO member’s rights as shareholders’ rights in Limited Liability Companies and Joint-stock Companies | Tokenization of shareholders’ rights of limited liability companies, etc. | Free distribution of the shareholders’ rights is not allowed under corporate law, etc. |
Sales by a third party for an issuer need a Type I license. A Type II license is necessary for the self-offering of a limited liability company. No license is required in the case of self-offering of a joint-stock company. In the case of solicitation of 50 or more people, there needs to be a submission of a registration statement regarding securities, etc. |
No regulation |
DAO member’s rights (with dividends), not including shareholders’ rights | TK investment, partnership investment, tokenization of rights that are difficult to classify into prescribed legal forms, etc. | Unregulated |
Sales by a third party for an issuer need a Type I license. Self-offering needs a Type II license. In the case of solicitation of 50 or more people, there needs to be a submission of a registration statement regarding securities, etc. |
No regulation (Possibility of Investment Management Business license in the case of securities investment) |
On the other hand, a DAO without dividends, etc., is also possible. Its regulations are as follows:
Tokens/NFT | Free Token Distribution | Sale of Tokens | Investment Management(Assuming no dividend) |
Utility Tokens | No regulation | Crypto Asset Exchange Business regulation | No regulation |
NFT | No regulation | No regulation | No regulation |
With respect to possible legal forms for DAOs, the following comparisons can be made:
Status | Legal Form | Limited Liability | Is it possible to distribute? | Avoid Double Taxation | Others, Comprehensive Evaluation |
No Legal Entity Status | Association without rights +Tokens with unclear rights | 〇? | 〇 | × |
△~〇 High degree of freedom. A good scheme if there is no problem with double taxation. |
Civil Law Partnership + Partnership Equity Token | × | 〇 | 〇 |
△~〇 High degree of freedom. A good scheme if there is no problem with limited liability. |
|
Investment Business Limited Liability Partnership + partnership Equity Token | 〇 | 〇 | 〇 |
× Although other points are reasonable, there are restrictions on investment destinations and businesses, such as not being able to purchase NFTs. It’s usually hard to use this scheme as DAO. |
|
Limited Liability Partnership + Partnership Equity Token | 〇 | 〇 | 〇 |
× There are valid points; however, to use as a DAO is problematic because of the need to register the name of the union member. |
|
DAO has Legal Entity Status |
Corporation (*1) + Tokenization of Anonymous Partnership (e.g., TK-GK scheme) |
〇 | 〇 | 〇 | △ It is necessary to operate in accordance with the Companies Act and the General Incorporated Associations Act. It should be noted that TK holders do not have the right to instruction. The good point is that there is no double taxation and limited liability. |
Corporation (*1) + Token with unknown rights |
〇? | 〇 | × | △ ~ 〇 It is necessary to operate under the Companies Act and the General Incorporated Associations Act. Besides that, it has a high degree of freedom and is a good scheme if you don’t mind the double taxation problem. | |
Corporations (*1) + Tokenization of shareholders rights (*2) |
〇 | 〇(× For general incorporated associations) | × |
× Is there a low degree of freedom due to the need to operate per the Companies Act and the General Incorporated Associations Act? For example, it is necessary to manage members as shareholders. |
*1 Legal entities include limited liability companies, stock companies, and general incorporated associations. LLCs are generally easier to establish and operate than joint-stock companies. If you want to have a more public image, use a general incorporated association.
*2 Membership rights of a limited liability company, stocks of a stock company, membership rights of a general incorporated association.
2.3 Tokenization of Rights
Tokenization of rights of funds or partnership, where there is an investment of funds (including money and crypto assets), investment management, dividends, or redemption of the principal of more than 100%, would be broadly considered a collective investment scheme (fund) under Japanese law. Below is the summary of the Definition of a Collective Investment Scheme.
Summary of Definition of Collective Investment Scheme |
Rights that satisfy the following (i) to (iv) |
(i) Rights under a partnership agreement as defined in Article 667(1) of the Civil Code, a silent partnership agreement as defined in Article 535 of the Commercial Code, an investment limited partnership agreement as defined in Article 3(1) of the Act on Limited Liability Partnership Agreement for Investment Business, or a limited liability partnership agreement as defined in Article 3(1) of the Act on Limited Liability Partnership Agreement for Investment |
(ii) The existence of a business (the “Invested Business”) in which money (including cryptographic assets) contributed or contributed by the person who has such rights (the “Investor”) is allocated to the Invested Business; |
(iii) The investors are entitled to receive dividends of profit generated from the invested business or distribution of assets related to the invested business; |
(iv) There are no exceptional circumstances, such as all investors being constantly involved in the business. |
The revised Financial Instruments and Exchange Act, which came into effect on May 1, 2020, created the legal concept of Electronic Record Transfer Rights. The rights of tokenized collective investment schemes usually fall under the Electronic Record Transfer Rights below.
Outline of Definition of Electronically Recorded Transfer Rights |
Rights that satisfy the following (i) to (iii) but exclude (iv) (Article 2, Paragraph 3 of the FIEA): |
(i) Rights listed in each item of Article 2, Paragraph 2 of the FIEA (funds, trust beneficiary rights, members’ rights of general partnerships, limited partnerships, limited liability partnerships, etc.); |
(ii) When they are expressed in property values that can be transferred through an electronic data processing system; |
(iii) When recorded in electronic devices or other objects by electronic means; |
(iv) Cases provided in the Cabinet Office Ordinance have considered the nature of distribution and other circumstances. |
The sale of this electronic record transfer right requires a Type 1 Financial Business registration. If soliciting more than 50 people, it will be a public offering (Article 2, Paragraph 3 of the Financial Instruments and Exchange Act), and a securities registration statement must be submitted based on Article 5 of the FIEA.
If the sale is limited to qualified institutional investors or wealthy people of 49 or less, and even if there is resale, there are technical restrictions so that other people cannot become DAO token holders.
When the Investment DAO is formed, it can be sold in such a limited form at first, and after it grows, it can be sold to the general public while complying with increased regulations.
2.4 Tokenization of Company Membership Rights and Financial Registration Regulations
Regarding tokenization of company membership rights, a Type I FIBO license is necessary when a third party sells the rights, and Type II is essential in the case of self-solicitation. In the case of tokenization of company membership rights (shareholders rights) of a joint stock company, a Type I license is necessary in the case of solicitation by a third party, and no license is required in the case of self-solicitation.
2.5 Regulations on Public Offerings, etc.
If any of the following applies, it becomes a public offering. In principle, it is necessary to submit a securities registration statement.
(i) When soliciting the acquisition of securities from 50 or more persons (excluding Qualified Institutional Investors (QII) in the case there are restrictions on resale other than QII); (ii) When it does not fall under any of the following categories: Private Placement for QII, Private Placement for Professional Investors, and Private Placement for Small Groups. |
2.6 Financial Regulations for DAOs without Dividends
If DAOs have no dividends, etc., they are not considered securities, but different financial regulations may apply.
Tokens/NFTs | Free token distribution | Token Sales | Investment Management (Assumption without dividends) |
Utility Tokens | Unregulated | Crypto Asset Exchange Business License | Unregulated |
NFTs | Unregulated | Unregulated | Unregulated |
Disclaimer
The content of this article has not been confirmed by the relevant authorities or organizations mentioned in the article but merely reflects a reasonable interpretation of their statements. The interpretation of the laws and regulations reflects our current understanding and may therefore change in the future. This article does not recommend investment in DAO. This article provides merely a summary for discussion purposes. If you need legal advice on a specific topic, please feel free to contact us.