AI Rivals Hold Private Safety Talks Amid Antitrust Scrutiny
OpenAI, Anthropic, and Google DeepMind have been in discussions about AI safety for several weeks, according to OpenAI's policy chief. The talks come amid calls for industry coordination to mitigate catastrophic risks, though antitrust concerns have been raised. OpenAI supports a provision requiring independent verification of frontier models.
The private discussions between OpenAI, Anthropic, and Google DeepMind reportedly extend beyond safety protocols to the creation of a shared industry standards body. OpenAI's policy chief confirmed support for a legislative measure that would mandate independent verification organizations inside frontier labs. This aligns with earlier hints from Sam Altman about private coordination with other AI leaders.
The talks occur against a backdrop of political division. President Trump and advisor David Sacks have publicly downplayed existential risks, while industry leaders like Hassabis have urged government oversight. Amodei's proposal for a narrow antitrust waiver was rejected by Lehane, who believes current coordination doesn't require one, though legal scrutiny remains a possibility given the competitive stakes.
The private coordination among leading AI labs could reshape how frontier models are governed, potentially reducing catastrophic risks through shared safety standards. However, it may also concentrate decision-making power in a handful of companies, raising concerns about competition and accountability. If antitrust scrutiny forces these talks to halt, safety progress could slow, leaving the public exposed to faster, less-vetted deployments. Independent verification, if adopted, could increase transparency, but may also create barriers for smaller players.