Anthropic, OpenAI and Google discuss forming AI safety standards body

The three leading developers of frontier artificial intelligence — Anthropic, OpenAI and Google Alphabet Inc. ↗ 1.53% DeepMind — have been holding behind-the-scenes discussions about forming a joint industry standards body for AI, according to a report by The Information published Sunday. The talks have been underway since at least July, predating Anthropic CEO Dario Amodei's public call on Saturday for the industry to slow down AI development.theinformation+1

A Push for Shared Safety Protocols

The proposed organization would establish shared protocols for technical testing and pre-release auditing of frontier AI models. Sam Altman, Demis Hassabis and Amodei have each publicly called for U.S.-led collaboration on AI safety standards, including international forums for testing and risk analysis.cryptobriefing+1

The three labs still disagree on key details. Anthropic has leaned toward government partnership, with its red team lead publicly advocating in July 2026 for standards developed alongside government bodies. OpenAI has emphasized voluntary industry rules, publishing a blog post in September committing to advancing standards regardless of government intervention. The discussions build on earlier coordination through the Frontier Model Forum, a consortium that also includes Microsoft ↗ 0.65%.aiweekly+1

Amodei's Essay Sets the Stage

Amodei's roughly 3,800-word essay, published Saturday, provided the public backdrop to the private talks. He warned that AI could be capable within six to 12 months of leading automated swarms that could "take over the entire internet," among other risks. "I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong," Amodei wrote.latimes

His three-step plan calls for embedding independent evaluators inside AI labs, coordinating common safety standards among companies in democratic countries, and pursuing global coordination — including with authoritarian governments. Anthropic has already committed unilaterally to the first step, offering third-party evaluators desks in its offices, access badges and company laptops.techcrunch+2

The China Question

The proposal's most contentious element may be international coordination. Amodei acknowledged on Sunday that China represents the "toughest dilemma" — the risk that slowing U.S. development simply hands an advantage to competitors who do not follow the same rules. The Future of Life Institute's AI Safety Index from summer 2026 underscored broader concerns about the industry's safety record, rating Anthropic highest among the three companies at C+, followed by OpenAI at C and Google DeepMind at C.cnbc+1