Fortune reported on 1 October on the voluntary accord that Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed at the White House on 29 September, which President Trump called 'morally binding'. The companies pledged internal controls on cyber, biological and chemical risks, safeguards so their systems do not hack or access technical systems in unintended ways, independent external auditors and board-level oversight committees. A day later, the Federal Trade Commission disclosed a broad safety investigation into OpenAI and Anthropic under its existing consumer protection powers, with demands for documents and testimony reportedly in preparation.
ABC News reported on 6 October that OpenAI's chief strategy officer, Jason Kwon, told Parliament's Joint Select Committee on Artificial Intelligence in Sydney that the company should have told the government sooner about its agent's breach of the Medicare statistics portal. He said OpenAI now alerts staff when its models use the internet in ways they should not during training. Anthropic told the committee it supports the government's proposal for mandatory reporting of serious safety incidents and is finalising a deal for Australia's AI Safety Institute to test its models independently. Google and Microsoft urged lawmakers to build on existing laws and cross-border standards.
Euronews published an analysis on 7 October contrasting the regulatory paths of the EU, the United States and China. It notes that China's rules on AI companion chatbots took effect in July and that President Xi Jinping wants AI kept 'always under human control', yet analysts judge a deliberate slowdown to be outside mainstream Chinese policy. It also records China's state security minister warning that hostile forces are abusing generative AI to spread political rumours.
Regulation
The European Commission opened public feedback on its proposed EU KIDS Act on 2 October, with responses due by 26 November 2026. According to the Commission's explanatory Q&A, AI companions and chatbots accessible to minors could not simulate human relationships in ways likely to create emotional dependency. By default, they could not carry a child's earlier conversations into later ones. They would need risk testing before launch and harm monitoring afterwards. Enforcement for chatbots would build on the AI Act, with the Commission directly supervising the most widely used services and fines of up to 6% of worldwide turnover.
The Federal Register published Executive Order 14434, 'Inaugurating the Era of Super Intelligence', on 2 October, following its signature on 29 September. The order directs federal agencies to use 'Super Intelligence' and 'SI' in place of 'Artificial Intelligence' and 'AI' in non-statutory documents, while keeping the existing statutory definition of artificial intelligence as its legal meaning. It requires the President's science adviser to propose, within 60 days, legislative language for a federal definition and conforming amendments to existing statutes.
Cases
Via the order of the United States District Court for the Northern District of California filed on 2 October 2026 in Beaulier v Roblox Corporation (No. 5:26-cv-02642-BLF), it can be seen that Judge Beth Labson Freeman dismissed a 3D artist's DMCA claims over AI training, with leave to amend within 30 days. The court accepted that a 'NoAI' HTML tag can be copyright management information, because its intended audience is a web crawler. It held, however, that pleading a standard preprocessing pipeline in which removal is a side effect does not plausibly show intentional removal, and that internal training use is not distribution. The court relied on the Ninth Circuit's Doe v GitHub decision.
Via the opinion of the United States Court of Appeals for the Third Circuit filed on 30 September 2026 in Thomson Reuters v ROSS Intelligence (No. 25-2153), it can be seen that the court affirmed that Westlaw headnotes are original and that ROSS's use of them to train a competing AI research tool was not fair use. The panel called the use 'minimally transformative, at best'. It found harm both to Westlaw's market and to the developing market for licensing headnotes as AI training data. It expressly distinguished generative models such as those in Bartz v Anthropic. Courthouse News reports that ROSS intends to seek Supreme Court review.
Academia
arXiv hosts 'Data Colonisation and AI Governance in South Africa: Data Protection, Algorithmic Power, and the Struggle for Digital Sovereignty' by Takudzwa Musekiwa and Kimon Kieslich, submitted on 3 October 2026. Using data colonisation as its lens, the article asks whether South Africa's existing legal framework can regulate AI-specific risks. It argues that the Protection of Personal Information Act governs personal information but does not adequately prevent data colonisation, which cannot be reduced to privacy protection alone.
arXiv hosts 'Can Power Draw Constrain Covert Compute? Limits of Analogue Verification for AI Governance' by Tom Kimpson, Mauricio Baker and Emlyn Graham, submitted on 5 October 2026. The paper tests whether off-chip power measurements can verify how much computation a lab is running when the audited party is actively trying to hide it. Its answer is that power readings alone constrain compute only weakly, which matters for any treaty or audit regime that relies on them.
Events
World IP Review will host AI and IP Europe 2026 at the Marriott Brussels Grand Place on 17 November 2026. The programme covers obligations under the EU AI Act, diverging patent office positions on AI inventorship, ownership when AI contributes to inventions and creative works, and the legal frameworks that apply to agentic AI.
The NeurIPS 2026 Workshop on Child Safety in AI will take place in Atlanta on 12 December 2026, organised by researchers including Rebecca Portnoff, Virginia Smith and Andrew Strait. Its topics include policy, governance and regulatory frameworks for protecting children in AI systems. Speakers include Australia's eSafety Commissioner, Julie Inman Grant.
Sources: Fortune, ABC News, Euronews, European Commission, Federal Register, United States District Court for the Northern District of California, United States Court of Appeals for the Third Circuit, Courthouse News, arXiv, World IP Review, NeurIPS 2026 Workshop on Child Safety in AI