Beating AI, according to a New York Times interview, revealed that former Anthropic employees mentioned that the CEO once suggested the company might gain power beyond that of a government. The employees had disagreements over AI security and development speed.
John Schulman advocates expanding public participation in AI alignment research and value formation. Mark Chen says there are early signs of recursive self-improvement in AI development, but it still cannot autonomously determine research directions or develop next-generation models.
Dongcha Beating AI reports that OpenAI CEO Sam Altman publicly opposed the social model in which AI evaluates human behavior and determines rewards, emphasizing that humans should retain control over the future.
Vitalik said that Kimi 3.1 excluding advanced offensive cyber capabilities is acceptable, and that society should harden systems and prepare before powerful cyberattacks arrive.
PANews, citing a CoinDesk commentary, said that six AI companies September 29 signed the White House Frontier AI agreement; the commentary argues that the agreement allows companies to choose their own auditors and scope of review, while lacking requirements for disclosure and penalties.
The Trump administration plans to transfer AI model testing institution CAISSI senior officials to the White House, but the transfers have not yet been completed and the individuals' identities have not been disclosed.
Microsoft CEO Satya Nadella said that superintelligence systems should not be regarded as incomprehensible black boxes and that enterprises should implement controls, audits, and human-oversight mechanisms outside the model.
Microsoft CEO Satya Nadella said that superintelligent systems should achieve observability, testability, and controllability through mechanisms including permission isolation, continuous verification, independent auditing, and incident disclosure.
Anthropic updated its usage policy, explicitly prohibiting users from repeatedly and gratuitously insulting or cruelly treating Claude and other models; the new rule will take effect on November 12.
Canadian Prime Minister Mark Carney on October 2 announced in Ottawa the establishment of the Prime Minister’s National AI Council, providing independent advice on Canada’s AI transformation and the “AI for All” national strategy.
Trump says he signed a document with “moral binding force” related to AI with executives from several technology companies and that the government is considering establishing an oversight committee; he also says he will rename AI it SI。
Trump said that Director of National Intelligence Jay Clayton is a suitable candidate for artificial intelligence director and that he plans to decide on the candidate within the next or days. The position's responsibilities and whether it will be formally established remain unclear.