Anthropic, Google and OpenAI have reportedly been holding regular discussions about creating a shared standards body for the artificial intelligence industry. According to reporting from The Information, representatives from the three companies have been meeting in a working group since July to discuss an organization focused on AI testing, auditing and shared safety standards. The talks are ongoing and reportedly continued as recently as the past week, although no formal organization has yet been announced.
The discussions involve representatives below the CEO level and appear to have begun before the companies’ leaders made several recent public calls for stronger coordination on frontier AI safety. The Information reports that the group has been examining how the industry could create common standards without waiting for the U.S. government to establish a regulator. Details such as membership requirements, governance, enforcement powers and the exact testing procedures remain unresolved.
Google DeepMind CEO Demis Hassabis publicly proposed a similar model in July, calling for an industry-funded standards organization that could evaluate powerful AI systems before deployment. His proposal drew comparisons with the Financial Industry Regulatory Authority, or FINRA, which operates as a self-regulatory organization for the U.S. securities industry while remaining under government oversight. Hassabis argued that an AI equivalent could initially rely on voluntary participation before potentially becoming part of a broader regulatory framework.
OpenAI CEO Sam Altman has also expressed support for an independent testing and auditing organization. According to The Information, Altman told employees at a recent company-wide meeting that major AI laboratories may need to establish such a body themselves if government support does not materialize. He reportedly acknowledged that the industry has not yet reached agreement on the idea, with some companies remaining opposed to a shared structure.
Anthropic CEO Dario Amodei has meanwhile called for stronger coordination between AI developers around testing and external evaluation. He has argued that leading laboratories should give qualified outside evaluators deeper access to frontier models and has also raised the possibility of narrowly tailored legal protections allowing competitors to discuss safety standards without creating antitrust concerns. Anthropic’s public safety framework already links increasingly capable models with progressively stronger safeguards and evaluation requirements.
The reported working group is therefore notable because it suggests cooperation between the three companies has moved beyond individual policy statements. Anthropic, Google DeepMind and OpenAI each maintain their own frontier safety frameworks, but those frameworks differ in terminology, testing thresholds and the commitments triggered when models reach particular capability levels. A shared standards organization could theoretically make model evaluations easier to compare across companies, although no common framework has been publicly agreed.
The idea is also likely to attract scrutiny over competition. Smaller AI companies could be affected if the largest model developers collectively establish testing requirements or deployment standards that are expensive to satisfy. Cohere CEO Aidan Gomez has already criticized the concept of the biggest laboratories coordinating on industry rules, highlighting the risk that safety standards could also become barriers to entry if governance is dominated by the market leaders.
Government involvement remains another unresolved question. A proposed White House effort to establish a similar AI standards organization reportedly lost momentum over the summer, while U.S. officials have generally favored voluntary industry action over a new federal licensing regime. Existing organizations including the National Institute of Standards and Technology already work on AI standards and model evaluation, meaning any industry-created body would also need to define how it interacts with government institutions and existing international standards groups.
OpenAI is already involved in several standards initiatives, including the Linux Foundation-hosted Appia Foundation, which is intended to turn existing AI frameworks into more practical and interoperable assessment criteria. The company also participates in standards work through organizations including NIST and international technical committees. A new industry body would therefore join an increasingly crowded landscape rather than starting AI standardization from scratch.
For now, the discussions between Anthropic, Google and OpenAI remain private negotiations rather than a formally announced organization. There is no confirmed name, launch date, membership structure or enforcement mechanism, and the three companies have not jointly published a charter. The significance lies in the fact that competing frontier AI developers are reportedly meeting regularly to explore common testing and safety rules, suggesting that industry-led coordination is becoming more concrete even as governments continue debating how directly they should regulate advanced AI.
