OpenAI, Anthropic, and Google discuss forming shared AI safety standards body

The CEOs of OpenAI, Anthropic, and Google DeepMind are in discussions to form a shared industry standards body to establish pre-release safety protocols for frontier AI models. While the initiative aims to resolve competitive tensions by standardizing risk assessments, critics warn that an industry-funded body faces severe conflicts of interest. Furthermore, high compliance costs could create steep market entry barriers for smaller startups, complicating an already crowded global regulatory landscape.

Proposed AI safety standards body

• ▪Anthropic CEO Dario Amodei proposed in September 2026 that independent evaluators like METR should visit frontier laboratories for extended periods to facilitate safety standard collaboration
• ▪OpenAI published a blog post in September 2026 emphasizing its commitment to advancing voluntary AI standards while simultaneously supporting specific legislation in California
• ▪Google DeepMind head Demis Hassabis proposed a U.S.-based Frontier AI Standards Body in July 2026, featuring voluntary technology submissions and a 30-day pre-release evaluation period
• ▪The CEOs of OpenAI, Anthropic, and Google DeepMind have held discussions about forming a collaborative industry standards body for artificial intelligence safety protocols
• ▪The proposed AI standards body would establish shared protocols for testing frontier models before public release, including independent evaluations and standardized risk assessments

Existing industry coordination efforts

• ▪The Frontier Model Forum was established by major AI companies, including Anthropic, Google, Microsoft, and OpenAI, to fund safety research and develop technical evaluations
• ▪The Appia Foundation was established in June 2026 with 13 members, including OpenAI, to translate general AI governance principles into specifications and compliance tests
• ▪The Future of Life Institute's AI Safety Index from summer 2026 graded Anthropic at C+ (2.66), OpenAI at C (2.28), and Google DeepMind at C (2.01)
• ▪The Agentic AI Foundation was created in December 2025 under the Linux Foundation to develop open standards and infrastructure for artificial intelligence agents

Independence and conflict-of-interest concerns

• ▪A GovAI projection indicates that 14 to 16 AI models from 2025 to 2028 could match the scale of the largest training runs, requiring continuous review capacity
• ▪An industry-financed AI regulator faces potential conflicts of interest resembling the issuer-pays credit-rating system that existed before the 2008 financial crisis

Competitive barriers for smaller labs

• ▪High compliance costs from industry-wide AI safety standards could create market entry barriers for smaller laboratories and startups lacking extensive computing resources
• ▪The OECD warned that high fixed expenses, concentrated infrastructure, and lack of computing resources already make it difficult for new companies to enter the AI market

International coordination with EU

• ▪Anthropic, Google, Microsoft, and OpenAI are signatories to the European Union's General-Purpose AI Code of Practice, which was introduced in July 2025
• ▪OpenAI Head of Global Affairs Chris Lehane urged the United States and China to support a global AI safety framework, comparing it to international nuclear cooperation

Debatable claims

• ▪An industry-financed AI standards body cannot provide independent safety evaluations
• ▪The US and China should establish a joint global AI safety framework
• ▪A shared AI standards body will unfairly block startup competition
• ▪Governments should regulate AI safety rather than relying on developer self-policing