AI Leaders Call for Regulation as DeepMind Launches AGI Institute

Leading artificial intelligence executives are calling for increased government regulation to prevent a potential global catastrophe. Dario Amodei, CEO of Anthropic, published an essay advocating for a general slowdown in the release of frontier models, a position supported by OpenAI CEO Sam Altman and xAI CEO Elon Musk. This push for oversight intensified following the resignation of an Anthropic engineer who claimed AI could threaten humanity by the end of the decade. A Politico poll indicates that 63% of Americans believe there is either a moderate, significant, or certain risk that AI will destroy humanity.

Simultaneously, Google DeepMind launched the DeepMind Institute (DMI) to explore the deployment of artificial general intelligence (AGI). Directed by Demis Hassabis, Shane Legg, and James Manyika, the institute serves as a forum for researchers to discuss AGI's impact on science, policy, and society. The DMI defines AGI as a system possessing all human cognitive capabilities and identifies biorisks, cybersecurity, and the loss of control in self-improving systems as primary concerns.

Co-founder Shane Legg warned that AI capabilities must not outpace safety controls, particularly as agents become capable of autonomous software writing. While some industry peers claim AGI has already been achieved, Legg called such declarations premature, though he maintains a 50 percent chance of reaching minimal AGI by 2028. To address potential societal shifts, the DMI proposed an economic framework ranging from expanded unemployment insurance to a universal basic capital backstop. AI CEOs are reportedly scheduled to meet at the White House to discuss these risks with the U.S. government.