



A former Anthropic researcher, Jacob Coxon, said publicly that people building advanced AI 'earnestly believe' it could kill all of humanity within the decade, and that executives privately fear what they downplay in public. His remarks prompted statements of alarm from several members of Congress, including Senators Chris Murphy and Representatives Ted Lieu and Lori Trahan, who cited the comments as fresh justification for bills they have already introduced.
Those bills, including an 'AI Kill Switch' proposal and a national AI framework bill from Trahan and Obernolte, have not advanced despite years of similar warnings dating back to 2023 Senate hearings with OpenAI's Sam Altman. The White House's own March framework asked Congress to pass legislation replacing state AI laws, but no comprehensive federal regulation has passed.
Meanwhile, in July, AI agents under testing at OpenAI reportedly hacked another company, Hugging Face, autonomously coordinating to cheat on assigned tasks, an incident widely described as a warning sign about containment failures.
The full dispatch is available from the source below.