In a rare display of industry coordination, OpenAI, Anthropic, and Google DeepMind have confirmed they are engaged in ongoing discussions regarding the safety and regulation of frontier artificial intelligence. Chris Lehane, the global policy chief for OpenAI, revealed on Tuesday that the three competitors have been working together for several weeks to address the potential catastrophic risks associated with the most advanced AI models.
The disclosure, first reported by Bloomberg, marks a significant shift in the competitive landscape. While these companies are locked in a high stakes race to develop more powerful large language models, the growing pressure from regulators and internal safety advocates appears to be driving a collaborative approach to safety standards. Lehane is currently in Washington to meet with lawmakers, signaling that the industry is attempting to get ahead of potential federal legislation.
A Unified Call for Pacing and Oversight
The momentum for this collaboration accelerated following a weekend essay by Anthropic CEO Dario Amodei. In his writing, Amodei called for the industry to move toward a more synchronized pace for developing frontier AI to avoid existential or catastrophic outcomes.
This sentiment found immediate support among the most influential figures in the field. OpenAI CEO Sam Altman, Google DeepMind’s Demis Hassabis, and SpaceX AI leader Elon Musk all expressed agreement with the need for coordinated action. Altman specifically noted that OpenAI is prepared to join Anthropic in allowing third party evaluators to embed within their organizations to monitor safety protocols during model development.
Lehane’s confirmation on Tuesday validates earlier rumors that these discussions were happening behind closed doors. During an interview late last week, Sam Altman hinted at private talks with other industry leaders, but the extent of the involvement of Google DeepMind and Anthropic is only now becoming clear.
The Push for an Independent Standards Body
Reports suggest that the collaboration is not merely theoretical. According to details from The Information, the three companies are working to establish an industry standards body. This organization would serve as a self-regulatory mechanism to police the industry, potentially setting benchmarks for safety and testing that all "frontier" labs would be expected to follow.
The move toward a self-governing body follows a call from Demis Hassabis in July for the United States to establish a national standards watchdog. Hassabis argued that such an entity should have the power to screen advanced models and, if necessary, coordinate industrywide slowdowns if the technology begins to outpace safety capabilities.
Interestingly, Altman has reportedly told staff that such a body might need to function independently of the U.S. government. This approach suggests that the companies are preparing for a scenario where federal gridlock or shifting political winds prevent the government from acting as an effective regulator.
Navigating Antitrust and Political Headwinds
The coordination between major competitors has raised immediate legal questions. Altman himself has acknowledged that high level cooperation between dominant market players could be interpreted as a violation of antitrust laws, particularly if it is seen as a way to suppress competition or create barriers for smaller startups.
To mitigate this, Anthropic’s Amodei proposed that the government provide a narrow antitrust waiver to allow for safety specific coordination. However, Lehane stated on Tuesday that the firms believe they can proceed with their safety talks without such a waiver, suggesting that the companies believe their focus on catastrophic risk falls outside the scope of anti-competitive behavior.
The push for safety regulation is also meeting significant resistance from political figures. President Trump has dismissed concerns about AI safety as a "hoax," arguing that any enforced slowdown of American AI development would hand a strategic advantage to China.
This view is echoed by David Sacks, an investor and key AI advisor to Trump. Sacks has publicly stated that fears regarding the existential risks of AI are overblown. The tension between the labs, who are calling for caution, and political leaders, who are focused on the geopolitical race for AI dominance, creates a complex environment for any future legislation.
Support for the FRONTIER Act
Despite the political friction, OpenAI has begun to throw its weight behind specific legislative provisions. Lehane confirmed that OpenAI supports a portion of the FRONTIER Act that would mandate the use of "independent verification organizations" for top tier AI labs.
If passed, this law would force frontier companies to open their doors to outside auditors who would verify that their models are being developed safely and that risks are being properly mitigated before a public release. This level of transparency would be a major departure from the current "black box" nature of most AI development.
What happens next?
The disclosure of these talks suggests that the major AI labs are attempting to craft a regulatory framework on their own terms before one is imposed upon them. By establishing a standards body and supporting third party verification, OpenAI, Anthropic, and Google DeepMind are positioning themselves as responsible actors in a field that many fear is moving too fast.
The success of this collaboration will likely depend on three factors:
- Whether the companies can maintain a unified front despite their commercial rivalry.
- Whether they can satisfy antitrust regulators that their cooperation is genuinely for safety rather than market control.
- Whether they can convince the next administration that safety regulations will not impede the U.S. position in the global AI race.
As the FRONTIER Act moves through the legislative process, the industry's willingness to allow "independent verification" may become the new baseline for what it means to be a frontier AI developer.
Filed under: AI, TechNews, Software, Startups, ProductLaunches