OpenAI to preview GPT-6 Cyber, launch security product: sources

Google, OpenAI, and Anthropic are reportedly teaming up to create an independent AI safety body by late 2026 or early 2027, according to a report from The Information. The collaboration aims to establish the Standards Authority for Frontier AI (SAFA), an organization that will operate separately from government oversight. SAFA will be tasked with setting practical benchmarks for the companies' public safety pledges and providing direct assistance to developers evaluating advanced models before they are released commercially.

Key figures involved in the initiative include former White House AI policy advisor Sriram Krishnan, who is reportedly being considered for the role of chief executive, and Arati Prabhakar, the former director of the White House Office of Science and Technology Policy. The three tech giants are also debating whether SAFA should conduct its own direct model assessments.

While the United States Center for AI Standards and Innovation already handles federal evaluations, industry insiders express concerns that the agency may not possess sufficient technical and financial resources to adequately inspect complex frontier systems. The push for a joint authority has gained momentum following weeks of industry dialogue, sparked by a proposal published on X by Sir Demis Hassabis, co-founder of Google DeepMind and Alphabet's chief scientist.

Hassabis advocated for a US-backed standards body modeled after a public-private partnership or a self-regulatory watchdog, drawing parallels to the Financial Industry Regulatory Authority (FINRA).

The collaboration comes amid growing concerns over potential catastrophic and security risks associated with AI development. Leadership at Google, Anthropic, and OpenAI has been under scrutiny due to the rapid release of frontier models. Anthropic CEO Dario Amodei recently called for an industrywide slowdown in the release of such models to allow safety mechanisms to keep pace.