AI View September 2026

Welcome to AI View, Simmons & Simmons' fortnightly round-up of key AI legislative, regulatory, and policy updates from around the world.

This edition brings you:

1. European Commission proposes AI chatbot and companion restrictions for minors in draft EU KIDS Act

2. China revises national AI safety governance framework to address risks from AI agents and embodied AI

3. Chinese regulator consults on draft rules restricting AI companion services for minors

4. US legislators introduce legislation to create new federal agency to ban artificial superintelligence and pause advanced AI development

5. California enacts laws requiring disclosure of AI-generated performers in advertising and explores new safety and security rules for frontier AI

6. Australia consults on national AI standards covering frontier model training and data centres

7. Hong Kong announces plans to appoint first AI Commissioner and publishes detailed regulatory strategy

1. European Commission proposes AI chatbot and companion restrictions for minors in draft EU KIDS Act

On 17 September 2026, the European Commission unveiled the draft EU KIDS Act (the Act), which introduces dedicated restrictions on AI chatbots and companion apps used by minors, alongside wider changes to social media age rules. The Act will need to be negotiated by the European Parliament and the 27 EU member states before it can enter into force.

Under the proposed rules, providers of AI companion chatbots (including those embedded in social media platforms) and similar apps would need to demonstrate that their products are safe for minors before launch, reversing the usual approach of regulators intervening only after harm has occurred. Key provisions of the draft Act include:

• A ban on chatbots simulating interpersonal relationships in a way that creates emotional dependency in users under 18.
• A ban on retaining memory of conversations with under-18 users across sessions.
• A prohibition on automatically activating or pinning an AI companion feature for a minor user; these must be turned off by default.
• A ban on profiling-based recommender algorithms, and new restrictions on other, non-profiling-based recommender systems.
• Prohibitions on addictive design features such as infinite scroll and engagement-maximising notifications.
• An increase in the age at which a minor can open a social media account without parental consent, from 13 to 15.

The rules would be subject to a carve-out for tools deemed to pose a low risk to minors, covering educational, scientific, and open-source development tools, and tools developed and operated by public authorities.

Penalties for non-compliance could reach 6% of a company's worldwide annual turnover. The draft Act now moves to the European Parliament and the Council of the EU for negotiation.

Read the Commission's announcement here and the draft Act here.

2. China revises national AI safety governance framework to address risks from AI agents and embodied AI

On 14 September 2026, China's National Technical Committee 260 on Cybersecurity (TC260), operating under the guidance of the Cyberspace Administration of China (CAC), released version 3.0 of the Artificial Intelligence Safety Governance Framework (the Framework). This is the third edition of China's national AI risk classification reference, succeeding the 2024 and 2025 versions, and is the first to introduce a dedicated risk management framework for agentic AI.

The Framework retains its existing structure of risk classification, technical countermeasures and comprehensive governance, but updates the risk taxonomy to reflect the shift of AI systems from answering questions to autonomously executing tasks.

Notably, it adds a standalone risk category for agentic AI, covering identity and permission abuse, reasoning and planning failures, tool-call execution risks and memory storage risks, supported by a dedicated Agentic AI Risk Management Framework and a new set of Fundamental Principles for Trustworthy AI. The taxonomy also introduces new categories for embodied AI risks and cybersecurity impact risks, including autonomous cyberattack behaviour.

Read the Framework here (English translation available from page 49).

3. Chinese regulator consults on draft rules restricting AI companion services for minors

On 18 September 2026, the Cyberspace Administration of China (CAC) published the draft Provisions of the State Council on Ensuring Minors' Healthy and Safe Use of the Internet (the Draft Provisions) for public consultation. The Draft Provisions consolidate and expand China's existing framework for protecting minors online, including obligations that already apply specifically to AI companion services.

Among the new rules, the Draft Provisions would:

• Introduce China's first outright prohibition on minors using stranger-based social networking services.
• Prohibit network service providers from offering minors any virtual relative, virtual partner or other form of virtual intimate relationship service.
• Require services that may affect minors' cognition — likely to capture a broad range of AI-driven products — to operate a “minors mode”, which must switch on automatically once a user is identified as underage, rather than being optional. Five methods for identifying minor users are set out, ranging from government ID checks to behavioural pattern recognition.
• Require providers to notify a guardian where a minor in minors mode shows signs of a crisis such as self-harm.

This builds on the CAC's Provisional Measures on the Administration of Human-like Interactive Artificial Intelligence Services, which took effect on 15 July 2026 and already prohibit AI companion services from providing virtual intimate relationships to users under 18, and require guardian consent for other anthropomorphic AI services offered to children under 14.

The consultation on the Draft Provisions is open until 17 October 2026.

Read the Draft Provisions here (in Chinese).

4. US legislators introduce legislation to create new federal agency to ban artificial superintelligence and pause advanced AI development

On 23 September 2026, US Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) introduced the Ban Artificial Superintelligence Act (the Bill), which would permanently ban the development of artificial superintelligence in the US and impose an immediate pause on advanced AI development until a new federal AI regulatory body is established with clear safety rules and model review processes.

Key elements of the Bill include:

• A prohibition on any person or entity developing or deploying artificial superintelligence, defined as AI that exceeds human cognitive performance and capabilities across most domains, or that has sufficient capabilities to destroy or disempower humanity, including by overthrowing the federal government.
• A pause on advanced AI development until a new federal AI regulatory body is operational and has established clear rules and model review processes for safe and secure AI development and deployment.
• The creation of a new cabinet-level Department of Artificial Intelligence, advised by leading scientists, to monitor frontier AI systems throughout their lifecycle for dangerous capabilities, supervise the removal of dangerous capabilities (such as the ability to subvert shutdown commands or conduct unauthorised cyberattacks), and supervise the destruction of any artificial superintelligence that is developed.
• Penalties for violating or circumventing the pause or the prohibition, including a "corporate death penalty" (legal procedure in which a government or court forces a corporation to dissolve). for entities and up to 20 years' imprisonment for individuals, comparable to existing penalties for unlawfully developing nuclear weapons.
• A US policy objective of pursuing international agreements, allied coordination and measures such as export controls to prevent the development of artificial superintelligence worldwide.

The Bill has been introduced in Congress and would need to pass both chambers before becoming law.

Read the press release here and the Bill text here.

5. California enacts laws requiring disclosure of AI-generated performers in advertising, and explores new safety and security rules for frontier AI

On 16 September 2026, California Governor Gavin Newsom signed Senate Bill 1050 (SB 1050) into law, adding a new Section 17610 to the California Business and Professions Code. From 1 January 2027, SB 1050 will require a clear and conspicuous disclosure whenever an advertisement prominently features a synthetic performer: a wholly AI-generated figure, voice or representation that is not recognisable as any identifiable real person. The law exempts advertisements for expressive works such as films, television and video games, and does not require advertising mediums (such as broadcasters or streaming platforms) to screen content proactively; their obligation to remove a non-compliant advertisement is only triggered once they are served with a court order.

Separately, on 18 September 2026, Governor Newsom signed a further executive order convening a panel of independent experts to develop recommendations for reinforcing and strengthening the state's AI safety and security laws, with recommendations due within two months (by approximately mid-November 2026). This order builds on legislation signed earlier in September 2026, including a framework for independent verification organisations to assess AI systems and a state registry for AI auditors.

Proposals under consideration reportedly include:

• Requiring frontier AI developers to host an independent verification organisation on-site to conduct regular audits,
• Requiring that safety frameworks, transparency reports and risk assessments already required under Californian law be independently verified rather than self-reported.
• The concept of an AI “kill switch”: an emergency mechanism to constrain or shut down a frontier AI system behaving in a dangerous or unintended way, although the Governor has acknowledged that the concept still needs to be defined by the expert panel.

Read SB 1050 here and the Governor's executive order here.

6. Australia consults on national AI standards covering frontier model training and data centres

On 17 September 2026, the Australian Government published a consultation paper, Getting it right: Building AI infrastructure that works for Australia, seeking feedback on proposed national AI standards.

The proposed standards cover two related but distinct areas:

First, they would set mandatory minimum requirements for large data centres, covering energy use, water use, cost impacts on consumers, community engagement, siting and workforce impacts.
Second, they would set conditions for frontier AI model training conducted in Australia, addressing how such training supports national security and safety, local industries and skills, research and innovation, and the interests of Australian creators, media and other rightsholders.

Submissions on the consultation paper are open until 9 October 2026, alongside targeted stakeholder engagement.

Read the consultation paper here.

7. Hong Kong announces plans to appoint first AI Commissioner and publishes detailed regulatory strategy

On 16 September 2026, Hong Kong's Chief Executive delivered the 2026 Policy Address, setting out an expanded AI governance and innovation agenda. Among the measures announced, the Government will create the post of Commissioner for AI under the Digital Policy Office, tasked with coordinating AI-related work across bureaux under the direction of the Chief Secretary for Administration.

The Policy Address sets out a seven-part AI risk governance strategy, covering:

• Combating the use of AI in criminal activity, including studying legislation on deepfakes and consulting on powers to remove AI-generated fraudulent content.
• Enhancing protection for minors, including a dedicated task force reviewing the risks of minors' social media and AI use.
• Promoting market assessment of AI ethical risks in public procurement.
• Supporting industry-led development of sector-specific AI application and governance frameworks.
• Examining legal liability for AI-related accidents.
• Introducing AI agent safety-management guidelines, to be issued next year.
• Assessing the impact of AI adoption on the labour market.

Read the Policy Address here.