Science & Tech · AI · 18 hrs ago
Independent AI safety evaluators gain prominence as companies seek oversight
Independent groups that test AI systems for capabilities and risks are gaining a bigger role as advanced models develop quickly.
Anthropic and OpenAI are among the companies seeking help from evaluators such as METR, Apollo Research and Transluce.
The push comes as the United States has not moved toward federal AI regulation, leaving companies to shape much of their own oversight.
US President Donald Trump has backed a voluntary approach that encourages companies to work with independent evaluators.
But it is not yet clear who will pay for the evaluations, what access the groups will receive or how they will report their findings.
OpenAI fired three employees over what it said were policy violations involving sensitive information; two of them say their contact with outside evaluators was behind their dismissals, a claim the company disputes.
OpenAI says it is finalizing contracts with third-party safety assessors and plans to announce details.
Independent AI safety evaluators are gaining prominence as Anthropic and OpenAI seek help assessing their models.
The evaluators, including METR, Apollo Research and Transluce, assess AI capabilities and risks and flag harmful behavior.
Anthropic CEO Dario Amodei has pledged to embed independent evaluators at his company, and OpenAI CEO Sam Altman endorsed the idea.
Questions remain about how evaluators will be funded, what access they will have and how they will report their findings.
A dispute at OpenAI has highlighted tensions over employees’ communications with outside evaluators.
- Who
- Anthropic and OpenAI are seeking support from independent AI evaluators, including METR, Apollo Research and Transluce.
- What
- Independent groups assess AI models’ capabilities and risks. Companies are considering bringing evaluators into their safety work.
- When
- The article was published on October 11, 2026; it describes several developments from the previous month and week.
- Where
- The story concerns the AI industry, including companies in the United States. Anthropic’s pledge was discussed in relation to its company.
- Why
- The companies are seeking oversight of advanced AI models and their risks; the article says evaluators have gained importance in the absence of a federal push for regulation.
OpenAI
Mikita Balesni and Tomek Korbak
Reason for dismissals
OpenAI
OpenAI said three employees were fired for violating its policies on accessing and handling sensitive company information.
Mikita Balesni and Tomek Korbak
Balesni and Korbak said they believed they were dismissed because of how they communicated with third-party evaluators.
Who is paying for these companies to do their work? How are they going to support them? You need an ecosystem, you need a viable business model for this.
tremendous self-policing
partner with an independent external auditor or evaluator
Trump presented a voluntary accord encouraging AI companies to partner with an independent external auditor or evaluator.
Amodei pledged to embed independent evaluators at Anthropic, and Altman endorsed the idea.
Amodei published an essay calling for a slower pace in advanced model development.
OpenAI fired three employees, citing violations of its policies on accessing and handling sensitive company information.
Balesni posted on X that former colleagues were confused and afraid to speak about safety and third-party evaluators.
OpenAI said it was actively finalizing contracts with third-party safety assessors.
- Evaluator groups
- METR, Apollo Research and Transluce
- Anthropic CEO
- Dario Amodei
- OpenAI CEO
- Sam Altman
- Brown University professor
- Suresh Venkatasubramanian
- OpenAI employees who disputed the dismissal explanation
- Mikita Balesni and Tomek Korbak
Sources
Related news
Anthropic AI submits false murder tip to Philadelphia police
Nadella calls for human-controlled emergency brake on advanced AI
US verdict re-examined over AI-generated video of victim
Three-quarters of online stores score poorly on AI shopping readiness
TikTok CEO predicts AI-driven scientific breakthroughs within three years
Microsoft expands Copilot Notebooks with productivity and automation upgrades
Chinese factories are adopting AI agents and taking the approach abroad
European rules slow rollout of remotely controlled river ships
Pharma companies invest heavily as AI reshapes drug research
Anthropic bars repeated cruelty toward its AI models
OpenAI fires three safety researchers after dispute over sensitive information
