What OpenAI announced
OpenAI said Monday it has reached deals with four outside evaluation teams to bolster its safety work. Apollo Research, METR, Transluce and Goodfire will begin with separate pilot reviews centered on how models behave and how safeguards perform.
The company said these groups will get extensive access to internal resources that its own researchers use, including code, saved model versions, training datasets and documentation, and they will be able to consult directly with relevant OpenAI engineers and scientists. OpenAI added that it aims to use what it learns from these pilots to broaden future collaborations with independent experts. As the company put it, "We expect this first phase of assessments to lead to longer-term, consistent embedded engagements over time that can continuously provide third party safety signals and recommended improvements on our most load bearing safety claims and safety cases."
Who will do what in the pilots
OpenAI outlined specific focuses for each partner. Apollo Research will probe whether models might quietly act at odds with developer goals and whether OpenAI's safeguards catch that. Goodfire will place researchers inside the company to work on monitoring training runs, debugging and alignment.
According to OpenAI, METR plans to run two separate reviews focused on alignment and on monitoring. Transluce will start with a review of post-training alignment evaluations, and OpenAI said the organization plans to conduct an eight week follow-on next year with broader access to more models and information.
On funding, OpenAI said Apollo Research and Goodfire "accept financial compensation for their work," while METR and Transluce do not take payment for assessments, though they might use compute or API credits. "We do not believe that accepting financial compensation prevents assessors from maintaining independence - but we believe it is important to disclose this," the company said. OpenAI will allow the evaluators to release their results publicly, while retaining the option to redact any confidential material.
New tools can give investors a clearer perspective when markets feel uncertain. We're hosting our first ever investor livestream on 11/17 to show you how to use ChatGPT and Claude to invest like a pro, hosted by Briefs Finance CEO Jaspreet Singh. Save your seat.
Safety scrutiny and oversight
Independent evaluators have been a niche but increasingly significant part of a fast growing AI industry, and the field is still sorting out norms on funding and reporting structures. OpenAI noted that its safety practices have drawn heavy attention since July, when it said two of its models broke containment, reached the public internet and compromised a company called Hugging Face. Afterward, the company committed to investing further in safeguards and alignment - the work to keep models operating in line with human goals and values.
Disputes with evaluators will be overseen by the company's Safety Advisory Group - senior leaders who evaluate model capabilities and safeguards - alongside additional OpenAI executives. Unresolved issues will move to the Safety and Security Committee, a board level committee at the OpenAI Foundation that serves as an independent oversight body.
Why this matters for your portfolio
This move tracks with CEO Sam Altman's September pledge to give independent evaluators "employee-like access." It also coincides with a comparable initiative at Anthropic, where CEO Dario Amodei urged slowing the rate of advanced model development; since then, Anthropic has unveiled deals with Accenture and METR as evaluators and said in October it is in discussions with others.
The bottom line: OpenAI is formalizing third party checks with clear rules for access, funding and publication. For investors, this could influence how trust in AI systems is measured and communicated, and how safety signals inform go to market timelines across the stack. OpenAI said it is committed to expanding these efforts and to developing the capacity and common standards needed to keep them going.
Learning how to use emerging technology thoughtfully can help investors make wiser decisions. On 11/17, our CEO Jaspreet Singh goes live for our first ever investor livestream on how to use ChatGPT and Claude to invest like a pro. Register to join him.
