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.

AI x Dispute Resolution Webinar Series

Our AI x Dispute Resolution webinar series resumes, with upcoming sessions on AI Incident Response / Product Liability (1 October) and Evidence in AI Disputes (5 November).

The series has already explored AI disputes, regulatory enforcement, collective actions and negligence, providing practical insights for legal, risk and compliance teams navigating the rapidly evolving AI landscape.

Register for upcoming sessions here.

This edition brings you:

1. UK Human Rights Committee publishes report calling for AI legislation and oversight body

2. G20 countries agree ministerial statement on AI and emerging technologies

3. UK Government rules out broad ban on superintelligent AI but leaves the door open to targeted regulation

4. California enacts landmark child safety laws on social media and AI chatbots

5. US House panel advances bills on AI, supply chain security and sustainability mandates

6. UK Commission sets out blueprint for lifecycle regulation of AI in healthcare

1. UK Human Rights Committee publishes report calling for AI legislation and oversight body

On 14 September 2026, the UK Parliament’s Joint Committee on Human Rights (the Committee) published a report titled “Human Rights and the Regulation of AI” (the Report), calling for a new AI Bill to address the scale and seriousness of the human rights risks posed by AI systems.

Committee Chair Alex Sobel MP said that “nowhere in the world, including the UK, has a current legislative and regulatory approach to AI that is fit for purpose”, and called for new legislation to “establish a comprehensive set of protections that deal with the entire AI supply chain and its lifecycle”, alongside a single AI regulator with the power to set policy, monitor performance and enforce compliance. He added that regulation should remain proportionate, targeting the areas of highest risk, while ensuring individuals know when AI is used in decisions that affect them and have access to redress when things go wrong.

The Report finds that the UK’s existing legal and regulatory framework is fragmented, relies on harm-specific and sector-specific legislation, and applies to AI primarily at the point of deployment, leaving responsibility mainly on users rather than designers. Regulators currently lack the power to test or evaluate AI systems before release, or to block release where systems pose unacceptable risks, and engagement with the AI Security Institute remains voluntary. The Committee also highlights specific human rights concerns, including inadequate transparency around AI-assisted decision-making, insufficient mechanisms for individuals to challenge such decisions, and threats to the right to privacy.

To address these gaps, the Committee recommends a new AI Bill built around a risk-based approach, under which low-risk systems face lighter obligations while higher-risk systems and models are subject to more demanding requirements across the AI lifecycle. The Committee recommends that certain uses of AI incompatible with human rights, such as subliminal techniques and inappropriate use of profiling or biometric data, should be prohibited outright following public consultation, while other high-risk AI systems should require prior approval before deployment.

The Committee further recommends establishing a single, independent AI oversight body on a statutory footing, to act as a central point of contact for concerns about AI use and to monitor AI-related harms and risks.

Read the Report here.

2. G20 countries agree ministerial statement on AI and emerging technologies

On 2 September 2026, G20 innovation ministers adopted a ministerial statement on emerging technologies, with AI embedded across six pillars covering policy frameworks, public sector adoption, skills, intellectual property, standards and supply chains.

The statement is structured around “The Carolina Principles for Emerging Technologies” and “The G20 AI Prosperity Objectives”, which are intended to guide national approaches to technology governance while supporting voluntary international cooperation.

The pillars cover:

• Human centric, proportionate and adaptable policy frameworks: ministers warn that outdated or inflexible regulation can unintentionally restrict innovation. The Carolina Principles emphasise investment in foundational research and domestic capabilities and use of existing sector specific regulatory approaches, with any new rules focused on genuinely novel issues.
• Use of secure and trustworthy AI to improve public services, with members aiming to identify and pilot high value public sector use cases.
• Skilled workforce development, including upskilling, reskilling, lifelong learning and AI literacy initiatives, supported by the AI Prosperity Objectives.
• Intellectual property, with continued study of how existing IP rules apply to AI, and an emphasis on copyright, consent doctrines, limitations and exceptions, robust IP and trade secret protection, and fair remuneration.
• “AI for Standards and Standards for AI”, highlighting AI’s role in streamlining trade, supporting secure and trustworthy models, and encouraging the development and use of technical standards.
• Industrial innovation and supply chains, promoting AI enabled logistics and risk management, investment and modernised manufacturing, while addressing unnecessary market access barriers and supporting “data free flow with trust”.

Read the G20 Innovation Ministerial Statement here.

3. UK Government rules out broad ban on superintelligent AI but leaves the door open to targeted regulation

On 1 September 2026, the Artificial Superintelligence Security Bill (the Bill) was introduced in Parliament by Labour MP Alex Sobel. The Bill is positioned as a first step towards international rules on AI security, proposing to ban the development of “superintelligent” AI.

However, as a private member’s bill, it is considered unlikely to progress without Government support: the Government has indicated it will not back the bill, while leaving open the possibility of more narrowly focused measures to address national security risks from advanced systems.

The Bill would prohibit domestic development of highly capable AI systems, defined as those able to “neutralise, displace, circumvent, subvert, or render ineffective” relevant human authorities and harm UK security. It envisages establishing indicators to assess when AI systems are approaching risky capabilities and behaviours, including self preservation and self development.

A Government spokesperson stated that the proposed measures are not regarded as “the right approach”. They nevertheless acknowledged that “additional targeted interventions” may be required in future to address significant AI related national security risks, with any such steps expected to be closely tailored to specific threats and calibrated to minimise adverse impacts on economic growth.

The Bill comes against a backdrop of heightened concern about frontier AI. Recent evidence to Parliament’s National Security Strategy Committee highlighted perceived gaps in the UK’s protection against autonomous AI threats and challenges in maintaining meaningful human control over increasingly autonomous systems.

Read the Bill here.

4. California enacts landmark child safety laws on social media and AI chatbots

On 10 September 2026, California Governor Gavin Newsom signed 13 bipartisan bills introducing what the state describes as the strongest child safety chatbot and social media laws in the US. The laws are not yet in force: most provisions, including the ban on addictive social media features for under-16s, are due to take effect on 1 January 2027, with the companion chatbot safeguards under “Adam’s Law” following on 1 July 2027. The laws target addictive social media design, regulate AI companion chatbots used by children, strengthen privacy protections, and expand safeguards against AI enabled sexual exploitation and abuse.

A central element is “Adam’s Law”, which regulates companion chatbots. It requires robust protections around child-facing chatbots, including crisis protocols for suicidal ideation, parental controls, and notifications if a child disables safety settings, and mandates independent child safety audits and annual risk assessments. The reforms also prohibit social media platforms from offering “addictive” features to users under 16, such as autoplay and algorithmic feeds based on user history and profile.

The new laws expand the scope of child sexual exploitation offences to include digitally altered or AI generated material depicting a person under 18 engaged in sexual conduct and enhance privacy safeguards by restricting targeted advertising and regulating the use of children’s data in AI systems.

The package sits within a broader California strategy on AI and online safety, which has included requirements for platforms to publish policies on harmful content, measures against “sexually explicit digital identity theft” and deepfake pornography, AI transparency obligations for frontier model developers, and emerging frameworks for AI auditing and verification. Newsom has continued to call for robust national regulation to match the speed and scale of AI driven risks.

Read the Governor’s press release here.

5. US House panel advances bills on AI, supply chain security and sustainability mandates

On 1 September 2026, a US House subcommittee advanced the Open-Source AI Leadership Act as part of a slate of bills aimed at bolstering supply chain security, promoting domestic open source AI models and countering overseas sustainability requirements.

The Open Source AI Leadership Act would direct the Commerce Secretary to support adoption of domestic open source AI models and study risks from foreign adversary developed models such as China’s Kimi K3. While Democrat Frank Pallone backed US leadership in open source AI, he warned that the bill could give the Secretary “open ended authority” to remove barriers to adoption.

The subcommittee also advanced the Chip EQUIP Act, which would bar recipients of CHIPS Act funding for semiconductor R&D from acquiring manufacturing equipment from “adversarial” countries such as China, covering lithography, wafer slicing and other key tools. Chairman Gus Bilirakis framed the measure as part of efforts to “push back against the Chinese Communist Party” amid intensifying competition over AI enabling chips.

Both bills were forwarded by voice vote to the full House Energy and Commerce Committee and remain at an early stage. They still require approval by the full Committee before any vote by the full House, and would then need to pass the Senate and be signed by the President before becoming law.

Access the Open Source AI Leadership Act here and the Chip EQUIP Act here.

6. UK Commission sets out blueprint for lifecycle regulation of AI in healthcare

On 10 September 2026, the UK’s National Commission into the Regulation of AI in Healthcare published its recommendations for a future regulatory and assurance framework for AI in healthcare, aimed at making the NHS “one of the most AI enabled healthcare systems in the world” while keeping the system people centred rather than tech centred.

The Commission concludes that existing medical device rules are not well suited to software and AI enabled products that are frequently updated, context dependent and prone to performance drift, and calls for a shift from one off, pre market checks to proportionate, lifecycle based oversight.

The report is structured around three core principles:

• “proportionate lifecycle regulation”, with risk based, flexible routes to market, clearer qualification and classification of AI devices, support for staged authorisation, sandboxes and post market surveillance;
• “system wide responsibility and safe management”, with shared but clearly delineated responsibilities for manufacturers, healthcare providers, professionals, regulators and government, including an “AI readiness toolbox”, contractual allocation of risk controls and coordinated workforce training; and
• “trust, transparency and predictability”, emphasising improved patient information about AI use in care pathways, stronger public and professional engagement, clearer communication of safety concerns, and more navigable regulatory pathways for innovators.

The recommendations are grounded in a UK wide research and engagement programme involving over 12,000 people, 760 written responses to a call for evidence and inputs from more than 100 experts across four specialist working groups. The Commission stresses that regulation should enable earlier access to safe, effective and equitable AI technologies, while maintaining public trust and avoiding widening existing health inequalities. A cross government response will follow, setting out how the recommendations will be taken forward.

Read the Commission’s recommendations here.