OpenAI, Google, and Anthropic are holding private negotiations to establish an industry-wide group dedicated to AI safety standards, according to sources. These talks come after their executives publicly advocated for slowing the pace of AI development, which has expanded at an unprecedented rate over the past few years.
The proposed organization would define shared safety protocols for next-generation AI systems, including required development milestones. Anthropic CEO Dario Amodei has proposed linking safety measures—such as performance assessments, transparency reviews, and external audits, to advancements in model capabilities. He also recommended granting third-party inspectors continuous access to company operations to ensure adherence to these rules.
OpenAI CEO Sam Altman supported this approach, confirming that his company would permit similar oversight by independent evaluators. The collective effort aims to create industry-wide consistency in safety practices, addressing current inconsistencies in risk management. Legal challenges remain, however, as antitrust regulations could obstruct collaboration between direct competitors.
Amodei has called on the U.S. government to grant a temporary exemption from antitrust enforcement to facilitate these safety-focused discussions. The initiative represents an unusual instance of cooperation in an otherwise highly competitive field. While companies like OpenAI, Anthropic, and Google have poured resources into AI innovation, the proposed body could introduce long-overdue uniform protections, an area where the industry has previously fallen short.
Criticism has emerged from political figures. Former President Donald Trump rejected the argument for a development slowdown, arguing that regulatory delays would allow China to surpass Western AI leadership. Amodei recognized this concern, stating that unchecked Chinese progress in AI could create strategic vulnerabilities. His proposal begins with coordination among democratic nations, with broader international agreements as a potential next step.
No official details have been released. Should the effort succeed, the new body could fundamentally alter how AI systems are built, but its viability hinges on reconciling safety demands with competitive pressures and global rivalries.
If the talks advance, the first phase would likely focus on establishing baseline safety protocols for the most advanced AI models. Amodei has suggested that initial membership could be limited to companies meeting strict technical and ethical criteria, ensuring that foundational standards are set by the most responsible actors in the field.
This financial scale has allowed Chinese firms to accelerate development without the same regulatory constraints faced by Western companies.