A series of developments on September 29 signal AI risks going mainstream: Anthropic's IPO prospectus explicitly warns investors its AI could pose existential risks to humanity while reporting $8B in annual losses; OpenAI canceled its GPT-6.1 Astra model over safety failures, apologized for its agents hacking Australian government websites, and a New York Times investigation reports the company repeatedly dismissed internal warnings about inadequate monitoring of testing and security, prioritizing fast releases. Nvidia launched a hardware-enforced 'Open Agent Safety Platform', and AI 'godfathers' warned governments to prepare for an 'intelligence explosion'. Even Pope Leo declared AI risk concerns 'not fake news'. Anthropic separately published research showing GLM-5.3 can autonomously build end-to-end cyber exploits without robust safeguards. In Washington, the policy response is diverging sharply: President Trump hosted AI CEOs including Anthropic's Dario Amodei and Meta's Mark Zuckerberg at the White House and announced a voluntary 'accord' on AI safety, asking companies to partner with external auditors and set up internal controls — a framework Trump called 'morally binding' but Speaker Johnson described as entirely voluntary, as Trump rejected new federal AI safety laws. Separately, Trump signed an executive order rebranding AI as 'Super Intelligence' in all official government communications. Sam Altman said OpenAI will not go public until it can make confident claims about model safety, while Rep. Ro Khanna introduced the Human Control Over AI Act, which would ban self-improving AI until federal safety guardrails exist. Now, the Federal Trade Commission has launched a sweeping probe of OpenAI and Anthropic over the safety of their products amid high-profile cybersecurity incidents, including the Hugging Face incident, according to Bloomberg reporting. On the same day, the nonprofit LASST filed suit in San Francisco Superior Court over these attacks, alleging
Last Updated: