OpenAI, Anthropic, Google in talks on AI safety
▼ Summary
– OpenAI’s global policy chief Chris Lehane confirmed that the company has been collaborating with rivals Anthropic and Google DeepMind on AI safety measures for several weeks.
– This coordination follows an essay by Anthropic CEO Dario Amodei calling for industry-wide efforts to slow frontier AI development to avoid catastrophic risks, a stance supported by major tech leaders.
– While some executives acknowledge potential antitrust concerns regarding such collaboration, Lehane stated that the firms do not require a government waiver to proceed with their safety discussions.
– The companies are reportedly working to establish an independent standards body to monitor advanced models, despite President Trump dismissing safety concerns as a hoax and opposing regulatory slowdowns.
– OpenAI supports provisions in the FRONTIER Act that would mandate independent verification organizations to ensure safe model development within top frontier laboratories.
OpenAI, Anthropic, and Google DeepMind have been engaged in confidential discussions regarding artificial intelligence safety protocols for several weeks, according to Chris Lehane, OpenAI’s global policy chief. Speaking to reporters on Tuesday, Lehane confirmed the ongoing dialogue with its primary competitors, a development first highlighted by Bloomberg. Lehane is currently in Washington, D.C., collaborating with U.S. lawmakers to address the potential catastrophic risks posed by advanced AI systems.
This disclosure follows a significant essay published Saturday by Anthropic CEO Dario Amodei, which urged the technology sector to collectively decelerate the development of frontier AI models to mitigate existential threats. The call to action garnered support from major industry figures, including Sam Altman of OpenAI, Demis Hassabis of Google, and Elon Musk of SpaceX AI. Altman specifically committed to partnering with Anthropic by integrating third-party evaluators into their operations to monitor safety standards rigorously.
Industry Coordination and Legal Concerns
Lehane acknowledged that these three tech giants have been negotiating safety measures for weeks. However, such coordination raises legal questions. Critics and some executives, including Altman, have warned that collaborative efforts among rivals could potentially violate antitrust laws if they are deemed to suppress competition rather than enhance public safety. While Amodei’s essay suggested a narrow government waiver might be necessary to permit such cooperation, Lehane reportedly stated that the companies do not believe a waiver is required.
The confirmation aligns with previous reports suggesting closer ties between these entities. In a late-week interview with Fortune, Altman hinted at private discussions with other AI leaders. Additionally, The Information reported earlier this week that the firms were working together to establish an independent standards body for the industry. Altman reportedly informed staff that this initiative must proceed without reliance on U. S. government backing.
Regulatory Divides and Political Pushback
Google’s Hassabis previously advocated in July for the U. S. to create a new regulatory agency capable of acting as an AI watchdog. This proposed body would have the authority to screen the world’s most powerful models and coordinate industry-wide pauses if dangers escalated.
In stark contrast, the current political leadership remains skeptical of these safety narratives. President Trump has dismissed concerns about AI safety as a hoax, arguing against stricter regulations. He contends that slowing down development would inadvertently benefit China in the global technological race. His key AI advisor, David Sacks, reinforced this stance, describing fears of existential risk from AI as exaggerated. Sacks holds significant personal stakes in various parts of the industry as a longtime investor.
Despite the political opposition, OpenAI maintains its position on oversight. During the same Tuesday meeting, Lehane emphasized that OpenAI supports a provision within the FRONTIER Act. This legislation would mandate that top frontier laboratories allow “independent verification organizations” access to their facilities to ensure models are developed safely. TechCrunch has contacted Anthropic, Google, and OpenAI for further comment on these developments.
(Source: TechCrunch)
