AI Titans Unite: OpenAI, Anthropic, Google Secretly Discuss Future of AI Safety

Major artificial intelligence companies OpenAI, Anthropic, and Google DeepMind have been engaged in weeks-long discussions concerning AI safety, a collaboration confirmed by Chris Lehane, OpenAI’s global policy chief, on Tuesday. These talks, initially reported by Bloomberg, underscore a growing industry focus on addressing the potential catastrophic risks associated with advanced AI development. Lehane's presence in Washington further highlights the industry's engagement with U.S. lawmakers on these critical issues.

The impetus for this cooperative effort gained public traction following an essay published by Anthropic CEO Dario Amodei, which advocated for industry-wide collaboration to temper the pace of frontier AI development and proactively avoid severe risks. Amodei’s call to action garnered significant support from prominent AI leaders, including OpenAI’s Sam Altman, Google’s Demis Hassabis, and SpaceXAI’s Elon Musk. Notably, Altman committed to OpenAI joining Anthropic in embedding third-party evaluators within their operations to monitor safety protocols and development.

Lehane's confirmation on Tuesday revealed that these discussions had been underway for several weeks. This collaboration also aligns with earlier hints from Altman in an interview where he mentioned private discussions with other AI leaders. Furthermore, The Information reported that the three companies are working towards establishing a standards body for the AI industry. Amodei's essay suggested a specific government waiver to allow such safety-focused coordination without violating antitrust laws; however, Lehane indicated that the firms believe such a waiver is unnecessary.

Despite the shared goal of enhancing AI safety, the collaborative efforts raise potential antitrust concerns. Leaders like Sam Altman have acknowledged that intense coordination among dominant industry players could potentially be interpreted as an attempt to suppress competition, thus risking violations of antitrust law.

The push for industry standards and oversight extends beyond the immediate collaboration. Demis Hassabis, in July, called upon the U.S. to create a new standards body specifically designed as an AI watchdog. This proposed entity would possess the authority to screen advanced AI models globally and coordinate industry-wide slowdowns if increasing dangers necessitated such action.

However, not all stakeholders share the same concerns or support the proposed regulatory measures. President Trump has reportedly dismissed AI safety concerns as a 'hoax' and has expressed opposition to stricter regulations, arguing that any slowdown in AI development could give China a strategic advantage in the global race. His key AI advisor, David Sacks, an investor with significant stakes in the industry, echoed this sentiment, suggesting that fears of existential risks from AI are exaggerated.

Amidst these varied perspectives, OpenAI continues to advocate for specific legislative provisions. At the same Tuesday meeting, Lehane affirmed OpenAI’s support for a provision within the FRONTIER Act. This provision would mandate that leading frontier AI laboratories allow 'independent verification organizations' to access their companies, ensuring that AI models are developed and deployed safely and responsibly.