OpenAI, Anthropic and Google have been holding talks aimed at establishing a shared standards body focused on AI safety, according to multiple reports, as the industry’s leading labs face mounting pressure to coordinate on responsible development of increasingly powerful models.
The discussions are said to center on common protocols for evaluating model risk, sharing best practices around safety testing, and setting baseline guardrails that could apply across the industry rather than company by company. Details of the proposed body’s structure and authority have not been finalized publicly.
The talks follow a joint call last month, backed by OpenAI, Anthropic, Google and more than 100 other companies, urging coordinated action to guard against risks from advanced AI systems operating outside intended limits.
Industry watchers say the move reflects a broader shift toward voluntary self-regulation as governments around the world weigh their own AI legislation, with companies hoping to shape standards before regulators impose them. Critics, however, have questioned whether an industry-led body can provide sufficiently independent oversight given the competitive pressures driving rapid model releases.
None of the three companies has issued a formal joint statement on the talks so far.