Google, OpenAI, Anthropic Form Joint AI Safety Regulator

What Happened

Google, OpenAI, and Anthropic have jointly established a shared artificial intelligence safety regulatory body, according to announcements from the three companies. The organisation is structured to set and enforce safety standards across all three firms, marking the first time major competing AI developers have agreed to a common oversight mechanism.

Background

The three companies are among the largest developers of frontier AI systems in the world. OpenAI produces the GPT series of large language models and the ChatGPT platform. Google develops the Gemini family of models and deploys AI across its search, cloud, and productivity products. Anthropic, founded in 2021 by former OpenAI researchers, develops the Claude series of models and has made AI safety a central part of its stated corporate mission.

All three organisations have existing internal safety teams and have previously published voluntary safety commitments, including pledges made to the White House in 2023 and participation in frameworks developed through the AI Safety Institute in the United Kingdom and its counterpart in the United States. Those commitments were non-binding and self-reported.

The announcement follows a period of sustained public and governmental pressure on AI developers to accept external accountability structures. Lawmakers in the United States, the European Union, and the United Kingdom have each advanced or implemented legislation requiring varying degrees of transparency, risk assessment, and incident reporting from AI developers.

Structure of the New Body

Details of the joint regulator's governance structure, enforcement powers, staffing, and funding arrangements have not been fully disclosed in available reports. It is not yet confirmed whether the body will have authority to halt model deployments, require third-party audits, or issue binding directives to any of the three member organisations. The relationship between the new body and existing national AI safety agencies has also not been specified.

The formation of the body by the companies themselves, rather than by a government or intergovernmental organisation, means its legal standing and enforcement capacity will depend on the terms agreed to by the founding members rather than on statutory authority.

What It Means in Practice

The joint body, if it operates as described, would represent a shift from purely voluntary, company-level safety processes toward a shared, cross-organisation standard-setting function. Whether that standard-setting includes binding requirements, independent auditing, or public reporting obligations remains to be confirmed.

The participation of all three firms creates a structure that covers a significant portion of frontier AI development activity currently underway in the United States. Other major developers, including Meta and Mistral, are not listed among the founding members.

Prior industry self-regulatory efforts in technology sectors have drawn scrutiny from regulators and researchers who questioned their independence and enforcement capacity. The new body's ability to demonstrate operational independence from its founding members is likely to be a central question as further details emerge.

Regulatory Context

The European Union's AI Act, which came into force in 2024, imposes mandatory obligations on developers of high-risk and general-purpose AI systems, including transparency requirements and conformity assessments. In the United States, no comparable federal statute has passed, though executive orders and agency guidance have directed departments to develop AI risk frameworks. The United Kingdom has pursued a sector-led approach through its AI Safety Institute.

The formation of a joint industry body in this environment may bear on ongoing legislative debates about whether self-regulatory mechanisms can substitute for or complement statutory oversight regimes.

The three companies are expected to release additional details on the body's governance structure, membership criteria, and operating procedures in the coming weeks.