A wave of departures from OpenAI's AGI and ASI safety teams has intensified questions about whether the company is prepared for the systems it is trying to build. Former OpenAI safety researcher Daniel Kokotajlo told Fortune magazine that around half of the company's safety researchers have recently left, including several prominent figures.
Kokotajlo did not give a single explanation for every resignation. But he said the pattern appears consistent with his own concern: OpenAI is "fairly close" to developing artificial general intelligence, or AGI, and is not prepared to "handle all that entails."
Why the safety departures matter
The central issue is not simply that employees have left a major AI lab. It is that the departures are concentrated among people working on some of the highest-stakes questions in artificial intelligence: how to manage the risks of advanced systems, including possible superintelligent AI.
According to Kokotajlo, about 30 employees had been working on AGI safety issues, and around 16 remain. He described the exits as not a "coordinated thing" but as people "individually giving up."
The list of notable departures includes Jan Hendrik Kirchner, Collin Burns, Jeffrey Wu, Jonathan Uesato, Steven Bills, Yuri Burda, Todor Markov, and OpenAI co-founder John Schulman. The resignations of chief scientist Ilya Sutskever and Jan Leike were especially significant because they jointly led OpenAI's "superalignment" team, which focused on the safety of future AI systems.
OpenAI later disbanded that team. In the context described by Kokotajlo, that move has become part of a broader concern about whether safety work has enough room, authority, and independence inside the company.
Concerns about publishing AGI risk research
Kokotajlo also pointed to a "chilling effect" on researchers who try to publish work about AGI risks from within OpenAI. He said there has been an "increasing amount of influence by the communications and lobbying wings of OpenAI" over what is considered appropriate to publish.
That claim matters because safety research depends not only on technical work but also on whether researchers can openly describe risks, uncertainties, and disagreements. If internal publishing decisions are shaped by communications or lobbying concerns, the public debate around AGI safety may become narrower than the research discussion itself.
The source article also notes that the temporary firing of OpenAI CEO Sam Altman was linked to safety concerns. After Altman was reinstated, a law firm cleared him.
Taken together, these details describe a company under pressure from several directions: rapid AI development, internal disagreement about risk, public scrutiny, and policy debates over how advanced AI systems should be governed.
The SB 1047 dispute adds a policy dimension
The departures are also tied to a larger argument about regulation. Kokotajlo said he was disappointed, though not surprised, that OpenAI opposed California's SB 1047 bill, which aims to regulate risks from advanced AI systems.
He co-signed a letter to Governor Newsom criticizing OpenAI's position. The letter called the company's stance a betrayal of its original plans to assess AGI's long-term risks carefully enough to help shape regulations and laws.
"We joined OpenAI because we wanted to ensure the safety of the incredibly powerful AI systems the company is developing. But we resigned from OpenAI because we lost trust that it would safely, honestly, and responsibly develop its AI systems," the letter reads.
This is a direct challenge to OpenAI's credibility on AI safety. The former researchers are not saying that advanced AI should be abandoned. They are saying they no longer trust OpenAI to develop it in the way they believe is necessary.
Researchers are leaving OpenAI, not AI safety
One of the most important details is where some of these researchers went next. Leike and Schulman moved to safety research roles at Anthropic, an OpenAI competitor that supports SB 1407 with some reservations. Before leaving OpenAI, Schulman said he believed AGI would be possible in two to three years.
Sutskever took a different path and founded his own startup to develop safe superintelligent AI. That choice reinforces the same point: these researchers are not walking away from the field. They are choosing other places to work on the same broad problem.
This distinction is important for understanding the OpenAI safety researcher exits. If former employees had left AI entirely, the story might suggest deep skepticism about the technology itself. Instead, the source article describes people continuing to work on AI safety, AGI, and superintelligence outside OpenAI.
The result is a sharper question for the industry: if leading safety researchers still believe powerful AI systems are coming, but no longer see OpenAI as the right place to manage those risks, what does that say about the company's internal alignment between development speed, public accountability, and safety research?
What the departures signal now
The facts in the source point to a widening trust gap. Kokotajlo believes OpenAI is near AGI but underprepared. Around half of its AGI/ASI safety researchers have reportedly left. A major safety team was disbanded. Former employees have criticized OpenAI's position on AI regulation while continuing safety work elsewhere.
None of that proves what OpenAI will or will not achieve. But it does show that the debate over AGI safety is no longer abstract. It is shaping staffing, research publication, company strategy, and public policy positions at the center of the AI industry.
For readers watching the future of artificial intelligence, the key takeaway is straightforward: the argument is not only about whether AGI can be built. It is also about whether the organizations racing toward it can keep enough trust, expertise, and openness in place to manage the consequences.