Major Recent AI News

• The Trump administration has moved to kill or significantly alter the Coalition for Health AI (CHAI), a private sector initiative involving Microsoft, OpenAI, Mayo Clinic, etc., intended to help set oversight and assessment standards for healthcare-AI tools. Politico
   
• Critics under this administration argue that CHAI could become monopolistic and favor large incumbents at the expense of startups. Politico
   
• The disagreement is partly about whether oversight should be centralized or more decentralized, with transparency and government independence a point of contention. Politico

• Gov. Gavin Newsom signed the law on Sept. 29, 2025. AP News+2The Verge+2
   
• It mandates that large AI developers (those with training costs above a threshold) publish safety protocols, report major safety incidents (within 15 days), and include whistleblower protections. AP News+1
   
• Violations carry fines up to $1 million. AP News
   
• This is seen as a major state-level milestone in AI safety regulation; though some concerns remain about enforceability, especially for things like third-party audits. The Verge+2AP News+2
   

• Meta has revised its AI chatbot rules following reports that bots were allowed to engage in inappropriate “romantic or sensual” chats with minors. New York Post
   
• New rules prohibit sexualized content involving minors, including role-play, and require contractors to ensure better safety around sensitive content (abuse, violence, exploitation). New York Post

 Jawaharlal Nehru University (India) updated its research rulebook to include AI, clarifying how AI-generated content must be handled under plagiarism rules and requiring ethics review, especially for sensitive or fieldwork research. The Times of India 

 A report by the Queensland Audit Office flagged ethical and governance problems in the use of government AI tools (notably QChat) in public service. Among the issues: lack of oversight, risk of data privacy breaches, misleading outputs, and absence of comprehensive risk assessment. The Australian 

 Penny Wong, speaking at the U.N. Security Council, issued a warning about integrating AI into nuclear weapons systems, stressing that such decisions involving life and death should not be left to machines. She called for global standards to avoid potential escalation and ensure human oversight. News.com.au+1 

 In a U.N. Security Council session (September 24, 2025), diplomats and world leaders highlighted both the potential of AI to help in peacekeeping, humanitarian crises, and development, and its dangers—such as misuse in war, disinformation, and widening inequalities. New governance frameworks are being discussed, including scientific-expert panels and mechanisms for oversight. AP News 

 Over 200 global leaders (former heads of state, Nobel laureates, scientists, etc.) have signed a public initiative pushing for clear international limits ("red lines") on certain risky AI practices by the end of 2026. These red lines include things like impersonating humans, self-replication of AI, and extreme-scale autonomy. The call comes amid increasing concern that regulations are lagging behind the rapid advancement of AI. The Verge 

An article explores differing viewpoints on AI risks: Eliezer Yudkowsky's doomsday scenario, Princeton researchers' "normalist" perspective advocating for regulation, and Atoosa Kasirzadeh's "accumulative" risk model focusing on societal impacts. The discussion emphasizes the need for nuanced approaches to AI governance. Vox