AI Law Wiki News for September 22, 2026

Davey proposes international treaty and AI safety agency

Liberal Democrat leader Ed Davey plans to seek an international treaty pausing superintelligent AI development. City AM reported on September 21, 2026, ahead of his planned September 22 conference speech, that the proposal includes an international AI safety agency based in the United Kingdom and modelled on the International Atomic Energy Agency. Davey argues for international oversight of concentrated technological power and criticises the UK government’s newly announced AI partnership with the United States as insufficient. For the existing national approach, see United Kingdom policy on catastrophic AI risk.

Senators seek inspector-general investigation of military AI errors

On September 21, 2026, Senator Jack Reed renewed calls for an inspector-general investigation into reported errors involving AI in military and intelligence operations. Speaking to CNN, Reed said Congress was still awaiting detailed information and wanted findings to inform possible legislative safeguards. Reed’s office published a joint request with Senators Mark Warner and Chris Coons for unrestricted inspector-general access to examine reported targeting and intelligence failures, including a strike in Minab, Iran, and an aborted interdiction operation. The senators seek public transparency and stronger human oversight. See United States policy on catastrophic AI risk for the broader governance context.

British Columbia sues OpenAI over Tumbler Ridge shooting

British Columbia announced on September 21, 2026, that it had filed a federal action against Sam Altman and OpenAI entities in California over an alleged failure to report threats before the February Tumbler Ridge school shooting. The Canadian Press reports that the federal lawsuit also names Sam Altman and seeks damages and injunctive relief, including safeguards against violent uses of ChatGPT. The province and school district seek recovery of costs associated with the attack.

New York sets November registration for frontier AI developers

Governor Kathy Hochul announced on September 21, 2026, that New York will direct large frontier-AI developers to register in November ahead of January 2027 compliance with the New York RAISE Act. She appointed Marc Gilman as deputy director supporting implementation through the DIGIT office. Hochul also said officials would explore additional safeguards, including the feasibility of AI shutdown mechanisms, as they develop future legislative proposals.

UK outlines AI and autonomy partnership with United States

On September 22, 2026, Downing Street outlined planned UK–U.S. work on an AI and Autonomy partnership. The UK Ministry of Defence’s Rapid AI Delivery Taskforce and its U.S. counterpart would collaborate on capabilities supporting critical-infrastructure protection and deterrence. The Guardian also reported the planned partnership ahead of Burnham’s UN address. See United Kingdom policy on catastrophic AI risk for the wider governance context.

Bessent announces formal AI dialogue and planned Shenzhen meeting

On September 21, 2026, Treasury Secretary Scott Bessent told CNBC that the United States and China had formalized an AI dialogue and agreed to meet again, probably in about two months in Shenzhen. He said the countries want an incident communications line and discussions about shared risks, including uncontrollable agents and non-state actors. Reuters separately reported the planned follow-up meeting and talks on AI-safety communications protocols. This follows the September 20 notification proposal. See the U.S. and Chinese policy overviews.

Spain launches twelve-month IA360 roadmap

On September 21, 2026, Spain presented IA360, a twelve-month roadmap for responsible AI deployment. Prime Minister Pedro Sánchez said the plan would emphasize oversight and traceability, public safety, trust and protection of vulnerable people, and would pursue strategic technology projects, economic development, stronger AI governance and a social contract. He said the government would convene employers, unions, political parties and public authorities the following month to seek a national agreement on AI. RTVE reported that the government described the initiative as a framework for clear rules and responsible deployment.

Amazon amends Perplexity complaint over Comet cloud access

On September 21, 2026, Amazon filed a first amended complaint against Perplexity after the preliminary-injunction appeal returned the case to the district court. The pleading asserts claims under the federal Computer Fraud and Abuse Act, California’s Comprehensive Computer Data Access and Fraud Act, and California tortious-interference law. Amazon alleges that Comet for iOS copied users’ Amazon authentication cookies and session data to Perplexity’s cloud servers, which then used the credentials in a virtual browser to request password-protected Amazon pages. The complaint says the architecture operated from March 18 until at least May 11, 2026, when Perplexity temporarily disabled the feature. See Amazon.com Services, LLC v. Perplexity AI, Inc..

Guterres calls for multilateral AI risk management

In his September 22, 2026 General Assembly address, United Nations Secretary-General António Guterres called for a multilateral AI risk-management framework with independent oversight. He urged safeguards for children and human control over life-and-death decisions, placing AI accountability among the central challenges for international cooperation.

Illinois creates AI Cabinet to advise on safety and regulation

On September 22, 2026, Governor JB Pritzker signed Executive Order 2026-07 establishing an Illinois AI Cabinet to assess AI-related risks and advise state agencies on incident readiness, public-infrastructure safeguards and further regulation. Capitol News Illinois reports that the advisory group will examine data-center incentives and approvals, state procurement and potential accountability measures, and formulate proposals through 2027. Appointed members have yet to be announced. See Illinois AI Law for the state's existing AI legislation.

Maryland outlines AI protections for workers and children

On September 22, 2026, Governor Wes Moore announced a Maryland AI policy framework focused on protecting residents, workers and children. His administration said the AI Subcabinet would develop legislative recommendations for frontier-model companies, including safety testing, independent evaluations, whistleblower safeguards and incident reporting. State agencies plan guidance on how existing anti-discrimination rules apply to AI-based employment decisions, while the administration will seek legislation on likeness rights, algorithmic rent coordination and chatbot safeguards for minors. Patch reports that the framework is a roadmap: some measures can be pursued administratively, while others would require legislation. See Maryland AI Law for existing state measures.

References

Browse previous daily digests →