AI Law Wiki News for September 21, 2026
AI-generated text. This page was generated using artificial intelligence.
US proposes AI incident notifications in talks with China
After September 20, 2026 talks in New York, Treasury Secretary Scott Bessent said the United States had proposed an AI incident-notification mechanism for consideration at the forthcoming Trump–Xi summit. The proposal concerns incidents reaching national-security significance and greater bilateral transparency, according to AP and Reuters.[1][2] Xinhua’s account confirms that the officials discussed AI.[3] The development relates to U.S. and Chinese approaches to international AI safeguards.
Claims open in Apple’s Siri AI settlement
The notice campaign and claims period in Apple’s $250 million Siri advertising settlement opened on September 21, 2026. The administrator accepts claims through December 21, 2026; the final-approval hearing is scheduled for February 24, 2027. The class covers qualifying U.S. residents who purchased eligible iPhone 15 Pro or iPhone 16 models in the United States between June 10, 2024 and March 29, 2025, for purposes other than resale.[4][5]
UN scientific panel examines safeguards against loss of AI control
On September 21, 2026, the Independent International Scientific Panel on AI released an advance thematic brief using the OpenAI–Hugging Face incident to examine risks from agents pursuing goals inconsistent with human intentions. It surveys accountability, incident reporting, safety assessments and independent controls as options for decision-makers.[6][7]
Amazon blocks Meta’s Muse shopping agent
Amazon blocked Meta’s Muse from shopping on its site, invoking contractual conditions and raising concerns about undisclosed automated access, GeekWire reported on September 20, 2026. The dispute extends the platform-access debate surrounding Amazon’s litigation against Perplexity. Amazon said discussions with Meta were continuing.[8] Meta’s September 8 launch statement says Muse protects credentials from the agent’s view and requests user approval before purchases.[9]
OpenAI proposes international frontier AI standards
On September 21, 2026, OpenAI proposed a U.S.-led process for common frontier-AI standards, including capability evaluations, human oversight of recursive self-improvement and incident-reporting thresholds. Governments would determine how standards enter their legal systems.[10][11] The proposal builds on an existing international evaluation network: NIST reported its publication of consensus practices and open questions on February 13, 2026.[12] See United States policy on catastrophic AI risk.
EU proposes sustainability ratings for data centres
On September 21, 2026, the European Commission proposed a common rating scheme for data centres above 500 kW, assessing resource use and grid contributions. AFP reports that the labels would grade energy and water performance from A to G.[13][14] The proposal faces two months of Parliament and Council scrutiny, with first labels expected in 2027. A separate consultation addresses minimum performance standards. See EU regulatory coverage.
Finland and Norway lead call for international control of frontier AI
On September 21, 2026, Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre launched a joint appeal for stronger international oversight of frontier AI. Finland’s presidency reported support from 22 leaders representing 20 countries and the European Commission. The appeal calls for mandatory testing before deployment, independent evaluations, coordinated standards and shared reporting of serious safety incidents. It asks UN member states to explore an international institution that could set standards, enable verification and convene states when capability thresholds are crossed.[15][16][17] The appeal remains open to additional governments, putting international verification and human control on the diplomatic agenda. See international frameworks and United Nations.
California signs seven data-centre oversight laws
On September 21, 2026, Governor Gavin Newsom signed seven data-centre bills: AB 1577, AB 2383, AB 2469, AB 2619, SB 886, SB 887 and SB 1168. The package addresses resource disclosures, electricity costs and local oversight of AI infrastructure.[18] CalMatters groups the measures into electricity-cost allocation, water and resource reporting, and environmental-review requirements. It reports that industry representatives warn the package could discourage investment, while supporters seek protections for residents and utility customers.[19] See California’s data-centre package.
Georgia lawsuit challenges approvals for OpenAI data centre
On September 21, 2026, One Hundred Miles and five residents filed a Georgia superior-court action challenging the approval process for OpenAI’s Project Camellia. The complaint alleges four Open Meetings Act violations and an Open Records Act violation, and seeks to invalidate the tax-abatement memorandum and land-transfer documents and obtain withheld sale records.[20] The Georgia Virtue reports that the dispute concerns the authority’s notice, agenda, closed-session discussions and approval of the land sale.[21] See the case article and Georgia AI Law.
References
- ↑ Matt Sedensky, Associated Press, “Bessent: US proposes AI incident alert system in talks with China,” September 20, 2026.
- ↑ David Lawder and Antoni Slodkowski, Reuters via MarketScreener, “Bessent proposes US-China AI safety notifications in talks with Chinese vice premier,” updated September 20, 2026.
- ↑ Xinhua, “中美经贸磋商在美国纽约举行” (China–US economic and trade consultations held in New York), September 21, 2026, reporting September 20 talks.
- ↑ Verita, Smartphone AI Settlement, claims and key dates, reviewed September 21, 2026.
- ↑ David Joseppi, The Apple Post, “Apple’s $250 million Siri AI settlement claims are now open in the US,” September 21, 2026.
- ↑ Independent International Scientific Panel on AI, “AI Agents, Misalignment and the Risk of Losing Human Control,” advance unedited version 1, September 21, 2026, pp. 2, 4 and 14–16.
- ↑ Sophie Denar, Unite.AI, “UN AI Panel Invokes Precautionary Principle on Loss-of-Control Risk,” September 21, 2026.
- ↑ Todd Bishop, GeekWire, “Amazon blocks Meta’s Muse AI assistant in new standoff over agentic shopping,” September 20, 2026.
- ↑ Meta, “Introducing Muse,” September 8, 2026, security and privacy description.
- ↑ OpenAI, “Building standards for the next phase of AI,” September 21, 2026.
- ↑ Unite.AI, “OpenAI Proposes US-Led Global Technical Standards for Frontier AI,” September 21, 2026.
- ↑ NIST, “International Network for Advanced AI Measurement, Evaluation, and Science Publishes Consensus Areas on Practices for Automated Evaluations,” February 13, 2026.
- ↑ European Commission, “Making data centres energy efficient thanks to a new EU rating system,” September 21, 2026.
- ↑ Agence France-Presse via Tech Xplore, “EU proposes energy rating for data centers,” September 21, 2026.
- ↑ Office of the President of Finland, “A Call for Control of Frontier AI Models,” September 21, 2026.
- ↑ Office of the President of Finland, launch announcement, September 21, 2026.
- ↑ Nilas Johnsen and Erlend Ofte Arntsen, “Norge tar initiativ til internasjonal kontroll med AI,” VG, September 21, 2026.
- ↑ California Governor, data-centre bills signing announcement, September 21, 2026.
- ↑ Katherine Ortiz, CalMatters, “Newsom clamps down on California data centers as voters turn against the industry,” September 21, 2026.
- ↑ Complaint, One Hundred Miles Inc. et al. v. Effingham County Industrial Development Authority et al., SUCV2026000155, filed September 21, 2026, pp. 1, 7, 18–26.
- ↑ Jessica Szilagyi, The Georgia Virtue, “Second Project Camellia Lawsuit Takes Aim at Secrecy,” September 21, 2026.