OpenAI Announces Embedded AI Safety Assessors, but Researchers Question the Program's Reach
OpenAI says outside experts from four organizations will scrutinize its safety work from inside the company, but critics say the access falls well short of what CEO Sam Altman promised in September.
By Harper Foster ·

OpenAI said Monday that staff from four AI safety and research organizations will work inside the company as "embedded assessors," an initiative the company described as a way to "give people a clearer view of how labs are evaluating and managing [potential] risks."
Speaking at Semafor's Silicon Valley and the World conference on Monday, CEO Sam Altman said, "Getting an outside opinion on the safety work and the robustness of these systems is going to be increasingly important and valuable." The program follows his September pledge that independent evaluators would receive "employee-like access" to internal company systems, a commitment made after details of the Hugging Face attack carried out by OpenAI's autonomous systems shook the AI field. Worries about the technology's catastrophic risk captured public attention, and rival Anthropic said it would set up a similar program.
Lama Ahmad, OpenAI's frontier assurances lead, defended the effort on X, writing: "We are giving deep access to some of the most sensitive tools and models our researchers use..that are the most important for assessing key, relevant and timely safety claims."
Many AI safety researchers, however, say the announced details describe something narrower than what Altman vowed. "The assessors have very narrow scopes to their access, and only two of them seem to really be allowed to talk to employees," said Sydney Von Arx, founder of the AI safety research group Nightingale.
Apollo Research, one of the announced assessors, outlined its limited access to OpenAI employees: if employees who have not been whitelisted, or pre-approved, reach out to the embedded assessors, the assessors must loop OpenAI in before holding substantive conversations. Von Arx said the ability to interview employees matters. "They help you find your blind spots. They tell you what you're missing." She also pointed to another omission: "Nobody has access to internal Slack," the corporate messaging tool. "Imagine being an 'employee' and only being allowed to talk about company information with whitelisted coworkers and having no Slack."
In a statement to NPR, OpenAI said there are currently no standards for embedded evaluations. "We expect to deepen these relationships over time and share what we learn to help establish best practices for effective embedded assessments across the industry," spokesperson Gaby Raila said. The company also said assessors can ask to speak with employees who have not been whitelisted, and that a process exists for obtaining statements from OpenAI about things learned informally.
According to OpenAI, the third-party organizations will focus on various aspects of AI safety, some related to preventing models from engaging in undesirable behavior and others to catching bad behavior when it happens. The assessors will have access to some internal models, training processes and some OpenAI employees, and they will be able to publish their findings after OpenAI redacts confidential information.
Outside evaluations are not new to the industry. AI companies have previously brought in external researchers to assess model capabilities and identify safety loopholes the companies missed, work that has raised public awareness of the risks the technology poses. One recent report about Anthropic using AI to significantly speed up model development prompted concerns about how close top companies are to using AI to autonomously build even more powerful models, a stage known as recursive self-improvement. Some in the field worry the velocity of development at that point could have catastrophic consequences for humanity, and they are calling for new regulation against the practice.
OpenAI's commitment to embedded evaluations came under further question after three employees who worked on safety were fired and later said they worried the company was walking back its promise. One of them, Tomek Korbak, had been the technical point of contact for two outside evaluators that investigated the Hugging Face attack. "Embedded auditing will only fulfill its aims if the auditors really get employee-level access as opposed to being hired to do a specific job," he posted on X, adding: "Sadly, OpenAI makes it clear now that [Sam Altman's] September tweet about 'employee-like access' was just an awkward overpromise that the company wants to quietly walk back."
More News · National news · Report a business · Got a Story?






