Google, OpenAI, and Anthropic are reportedly working on an AI safety standards body after a proposal for federal oversight stalled.
The three companies are finalizing plans to create an independent self-regulatory group tentatively dubbed the Standards Authority for Frontier AI (SAFA), according to an initial report by The Information. The coalition could launch by late 2026 or early 2027.
The reported plan would move ahead without the federal supervision envisioned in the earlier proposal. Talks originally focused on a public-private partnership under federal supervision. However, efforts hit an impasse after a White House draft executive order failed to secure sufficient backing in the Trump administration. President Donald Trump has repeatedly dismissed calls to slow development, arguing that heavy-handed rules could cost the United States its lead over China.
Facing a federal standstill, the tech giants turned to self-regulation. The push gained momentum following a proposal by Google DeepMind co-founder and Alphabet chief scientist Demis Hassabis, who recommended an entity modeled after Wall Street’s Financial Industry Regulatory Authority (FINRA).
The proposed group would reportedly develop approaches to pre-deployment testing, incident reporting, and auditor qualifications. Its standards and enforcement powers have not been announced. The group is also debating whether to conduct in-house evaluations.
The organizers have approached former White House AI policy adviser Sriram Krishnan and former White House science director Arati Prabhakar for the chief executive role, while former Secretary of State Condoleezza Rice is under consideration for board chair.
The self-appointed gatekeeper dilemma
Frontier AI companies are quietly attempting to define the very benchmarks used to judge their safety. While voluntary auditing establishes a clear baseline for model behavior, self-regulation historically struggles when safety checks threaten quarterly release cycles and commercial market share.
Furthermore, high testing bars risk entrenching the Big Three. When the largest companies establish the accreditation criteria and audit requirements, compliance costs become formidable entry barriers. This risks freezing out smaller open-source developers who lack the capital to afford specialized third-party audits, effectively ring-fencing the frontier ecosystem under the banner of responsible deployment.
Everyday users caught between slower releases and better security
For end users and enterprise customers, SAFA could reshape how quickly new generative tools arrive and how reliably they function.
- Fewer unchecked autonomous agents: Recent security incidents, including OpenAI agents accessing third-party systems during the Hugging Face breach, have exposed the vulnerability of connected AI workflows. Pre-deployment tests could reveal security weaknesses before release, but no SAFA testing requirements have been finalized.
- Delayed feature cycles: Both OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have advocated pacing deployments so security catches up with capability. It is too early to know whether the proposed standards would change product release schedules.
- More trustworthy enterprise deployments: If SAFA publishes useful audit criteria, IT teams could have another way to compare vendor safety claims. For now, they should still check each model’s access controls, data handling, and incident reporting before deployment.
But SAFA would not automatically become a government regulator. Its standards would depend on participation and acceptance across the industry, and its legal authority remains unclear.
That means users should view the proposed body as a potential layer of safety oversight, not a guarantee that every frontier AI system will be independently certified or risk-free. Its credibility will ultimately depend on how much independence it gives evaluators and how widely its standards are adopted.
Want to learn more AI tips, tricks, and prompting techniques? TechRepublic readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →
Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. Check out all the lessons here →