Japan weighs mandatory AI safety duties and penalties | Magica
Japan considers binding AI safety duties, with key limits still undecided
Editorial Team
••📖6 min read
Japan is considering safety and reporting duties backed by enforceable orders for advanced AI. The proposal would strengthen its nonbinding framework, but capability thresholds, covered businesses and penalties remain unresolved.
Japan is considering mandatory safety measures and reporting, with penalties backing government orders.
The proposed framework would focus on specified risks in AI above a capability threshold that has yet to be defined.
Responsibility across developers and service providers, and enforcement against overseas firms, remain central choices.
Japan is considering mandatory safety measures and reporting for advanced AI, backed by government orders and penalties for disobeying them. The proposed powers would strengthen an AI law that currently relies on investigations and nonbinding guidance, but the government has yet to define the capabilities, risks and businesses they would cover.
The draft targets the intersection of higher-performance AI and specified risks for possible binding regulation. Both boundaries remain under discussion. Source: Japan Cabinet Office, page 5. Source: Japan Cabinet Office.
From guidance to enforceable orders
The proposal was presented at a closed meeting on November 11 of a working group examining AI-related legal systems under the Cabinet Office, the government department supporting the prime minister. A report citing a person familiar with the matter said the government was considering safety requirements and penalties as concerns grow about unpredictable advanced AI behavior and suspected AI misuse in data leaks and unauthorized access incidents.
The existing law is the Act on Promotion of Research and Development, and Utilization of AI-related Technology. Enacted in May 2025, it was fully brought into force on September 1, 2025. The Cabinet Office described its purpose as promoting innovation while addressing risks, citing both Japan's lag in AI development and use and public concerns about AI. That dual purpose matters to a review that continues to favor targeted duties over rules covering every AI system.
As the report describes, the law already allows investigations and guidance in cases of malicious conduct but contains no penalty provisions. The proposed change would give authorities a binding response when a business refuses to cooperate following an AI incident.
Google launched Gemini 4 Hydrogen and automatic model switching in Antigravity, claiming near-Carbon task success with less quota and faster completion. Lower token rates support the efficiency pitch, but reasoning settings, mixed competitive results and inconsistent benchmark labels leave the cost of completed work to be tested.
Editorial Team
The official discussion paper sets out a broader structure than punishment after an incident. Its draft framework would establish duties to take safety measures against specified risks, require businesses to report relevant information to the government, authorize recommendations to secure compliance and permit stronger orders backed by penalties when necessary.
These remain proposals for deliberation. The paper combines a summary of earlier working-group opinions with draft policy directions; it does not establish an enacted enforcement regime.
Two boundaries would determine the reach
The official paper places potential binding regulation at the intersection of two conditions: AI exceeding a certain performance level and particular risks judged to require AI-specific legislation. Other combinations would be handled through existing laws and guidelines.
Neither boundary is settled. Participants argued that capabilities should matter alongside computing resources when evaluating models. That distinction matters because a rule based only on the amount of computation would measure a development input, while the group's concern includes what a system can do and how it behaves.
Cyber risks and chemical, biological, radiological, nuclear and explosive threats feature in the discussion. Participants considered both deliberate misuse and malfunction, including unexpected behavior as AI becomes more agentic. They also identified sexual deepfakes and protection of young people as priorities. Those concerns should not be read as an agreed list of risks covered by the proposed penalties.
The paper repeatedly preserves a role for existing privacy and sector-specific laws. Its approach therefore depends on identifying gaps that require new AI duties, rather than treating every harm involving AI as a matter for the AI law.
The draft prioritizes consideration of duties on AI businesses while retaining a role for user governance and existing laws. Which business activities would carry duties remains unresolved. Source: Japan Cabinet Office, page 6. Source: Japan Cabinet Office.
Who controls the model—and who carries the duty?
The government still has to decide which businesses would be covered across research, foundation-model development and services sold to users. A separate account citing Mainichi's reporting said the scope under discussion includes major US technology companies. It described corrective orders as a means of dealing with uncooperative businesses when powerful AI poses serious security or safety threats.
Businesses also had a direct voice in the consultation. Six companies spoke to the group, according to the account of the meeting. Participants included Japanese telecoms group NTT and the Japan operations of Anthropic, developer of Claude, and OpenAI, developer of ChatGPT. Companies were asked about risks, their safety work and the rules they would support. Their participation does not establish agreement on the proposed obligations.
The official paper gives a practical reason why the allocation matters: downstream businesses using foundation models cannot inspect those models' internal workings. Participants called for responsibility across the supply chain and information to flow downstream. Duties imposed on a service provider would need to account for what that provider can actually observe or change.
Participants also identified development and experimental stages outside the reach of sector-specific regulation, and raised the difficulty of handling open-weight models whose developers are unknown. Both complicate a regime that needs a clearly identifiable business to report risks and comply with orders.
The draft directions prioritize duties on AI businesses while continuing to consider users' responsibilities. They retain a role for user governance and enforcement of existing criminal and industry laws.
Foreign enforcement is another unresolved constraint. Participants explicitly warned against a framework whose burdens fall disproportionately on Japanese firms because overseas providers are harder to reach. They also called for protection of trade secrets, predictable requirements and avoidance of overlapping reporting systems. These concerns put the design of information gathering alongside the severity of punishment.
Participants also called for stronger staffing and capabilities at Japan's AI Safety Institute, alongside better information sharing with the government. The institute was established in February 2024 by the Information-technology Promotion Agency to examine AI safety-evaluation methods. Its initial remit included work on evaluation standards and international cooperation. The working-group paper thus identifies evaluation capacity, as well as legal authority, as work still needed.
The next decision is the scope of compulsion
The working group aims to produce an overall outline, including possible legal revisions, by year-end. The account of the draft says the size of any penalty has not been set.
The official paper uses European and US state frameworks as reference points for risk management, reporting and enforcement. But it supplies no Japanese numerical capability threshold, penalty schedule or incident-reporting deadline. Importing those details from its overseas comparison would overstate how far Japan's proposal has advanced.
A further choice concerns timing. Participants questioned whether authorities should be able to demand reports when harm is feared, since they described the current law's Article 16 as premised on an infringement of rights or interests. Some also proposed emergency orders or direct penalties where public safety is at stake; those ideas are not settled provisions.
The resulting outline will need to show both which risks justify compulsory intervention and which business has the information and control to respond. Until those choices are made, the proposal establishes a direction toward enforceable duties, rather than a defined compliance regime for AI providers.
Google says its new fast AI model is reaching Search users worldwide, improving complex answers and source discovery. The rollout expands access to Gemini 4, but its effect on answer quality and website traffic remains unmeasured.
Four nonprofits are asking a California court to stop five AI developers from building or deploying more capable systems. The case seeks enforceable limits even as OpenAI has previously reported a training pause and is launching defensive AI tools.
Expo Simulators gives coding agents remote iOS and Android environments with recordings for reviewers. The launch extends Expo's cloud development services into verification, with per-minute charges and shared capacity limits.
Kleiner Perkins and Clearlake Capital led Mirendil’s $1 billion Series A. Its agent-led model-building experiment points to smaller research teams, but company benchmarks, substantial compute costs and a private beta qualify the promise of wider access.
Mount Thor raised $45 million in equity, alongside separate hardware financing, to expand Apple infrastructure for AI agents and model inference. Existing Mac clouds and workload-specific performance limits put its demand and cost claims to the test.
Salesforce Ventures led a Series C that values Q-CTRL above $1 billion. The funding backs GPS-free navigation and quantum computing software, while undisclosed revenue and contested performance benchmarks leave the commercial case open.
Adobe demonstrated 10 experimental creative tools and opened Deep Brush and Hands On for testing. The lineup explores more direct control over AI edits, but existing alternatives and unproven production performance limit claims of novelty.
House Foreign Affairs Committee Chairman Brian Mast says outgoing Nvidia chief Jensen Huang has agreed to questioning, with a date still under negotiation. Speaker Mike Johnson says support for limiting chip exports to China is growing as lawmakers examine allegations involving Huang and a cloud partner.
Leaked correspondence and a reconstructed publishing program expose John Mark Dougan’s use of AI to stock imitation news sites with pro-Russian material. The records illuminate funding and working methods, while his claims of influencing other AI systems remain unproven.
William Blair raised AMD to Outperform as server-chip prospects and valuation improved. AMD already expects revenue growth above 80% in 2027, but customer incentives, custom accelerators and weaker quarterly cash generation complicate the investment case.
Apple’s $299 first smart display earns praise for fast home controls and useful kitchen features, but reviewers found gaps in how it keeps family members’ messages separate. Its strongest advantages also depend on owning compatible Apple devices and accessories.
Donald Trump proposed Apple executive chairman Tim Cook to succeed Jensen Huang at Nvidia. Cook’s interest and the seriousness of the suggestion remain unclear, while finance chief Colette Kress has taken interim charge.
Norwest’s largest fund is a third bigger than its predecessor, giving the Wells Fargo-backed investor more capital for AI alongside healthcare, consumer and industrial businesses. The firm has not disclosed how much will go to AI.
Nebius raised its 2026 capital spending forecast to about $27 billion while steering revenue to the low end of guidance. Better customer prepayments support expansion, but delayed capacity and a $325 million non-recurring contribution complicate its quarterly revenue beat.
Megaspeed founder Alice Huang is reportedly living in Beijing, but her activities and the espionage allegations remain unresolved. Jensen Huang’s resignation changes Nvidia’s leadership without settling questions about her business connections and access to its chips.
Singapore says it barred former Megaspeed chief Alice Huang after she gave inaccurate information about her role, while finding insufficient evidence against her over server declarations. The wider investigation still depends on tracing shipments beyond Malaysia.
Nvidia’s longtime finance chief takes over from its co-founder as the board promises continuity, searches for a permanent CEO and faces unresolved questions over security and executive conduct.
Chicago Public Schools is seeking a way to preserve Google Search without its generative AI features as parents, educators and students press for stronger district rules. Google’s planned access changes turn an AI safety dispute into a test of who controls classroom technology.
Warby Parker scheduled November 12 preorders and holiday deliveries for its first Android XR glasses. Google’s backing combines development funding with a conditional equity investment, while the retailer expects only a modest initial sales contribution.
