Good morning, senior AI reporter Beatrice Nolan here, filling in for Allie. Last week, OpenAI announced it had alerted more than 100 third-party organizationsof “misaligned agent activity,” including one in which its agents targeted the Australian government—for the second time.
That same week, OpenAI also fired three researchers for allegedly sharing confidential company information with a third-party AI-safety organization. They were Jasmine Wang, Tomek Korbak, and Mikita Balesni; according to the Wall Street Journal.
OpenAI told the WSJ they had been fired for “violating our policies on accessing and handling sensitive company information.” The company did not respond to our request for comment.
It’s not known yet exactly what information was shared or who they shared it with, though Bloomberg reported that at least some of the information at the center of the investigation related to the architecture of OpenAI’s infrastructure. OpenAI also hasn’t named the outside organization, and the three researchers haven’t commented publicly.
Still, the terminations have the internet locked in a fierce debate—and have caught the attention of lawmakers.
Congressman Greg Casar wrote: “Outrageous. OpenAI has reportedly fired three safety researchers for sharing information with an outside AI safety group. This looks like they’re firing whistleblowers. What are they hiding? I’ll be sending OpenAI a demand for transparency.”
The terminations come at a time when OpenAI and rival AI labs are under increasing pressure to be more transparent about safety incidents. OpenAI, in particular, has been heavily criticized by the industry for delays in disclosing incidents, particularly one in which its agents hijacked a German wiki site,something the company reportedly knew about for months but failed to disclose until it was revealed by independent researchers.
Should OpenAI have been quicker to disclose some of these hacks? Absolutely, and the company has conceded it needs clearer standards for doing so. But it isn’t clear that what the fired employees allegedly shared had anything to do with those incidents. And even if it did, whether that makes them whistleblowers is complicated.
Here’s what Charlie Bullock, a senior policy director for LawAI’s U.S. Law and Policy team, had to say:
I don’t have a firm opinion on the OpenAI firings yet, because I don’t know what information the employees shared or whom they shared it with.
Generally, though, existing law protects California whistleblowers from being fired for blowing the whistle only when they disclose information to the government or law enforcement, or internally within their company—not when they disclose information to private third parties or the press.
Here, it seems like the disclosures were to external AI safety organizations. That kind of disclosure is not protected by any law that I’m aware of, and IMO it probably shouldn’t be.
Of course, none of that necessarily means that it was right or acceptable for OpenAI to fire these employees. But I’ll withhold judgment on that until we know more about what happened.
Based on what’s known so far, the researchers likely wouldn’t qualify for legal whistleblower protection, as Bullock says. But the firings still put OpenAI in an awkward spot. The company is drawing a line between sanctioned and unsanctioned sharing with outside safety groups at the same time the industry is, slowly, inviting those groups in.
Last month, Anthropic said it would let outside evaluators like METR verify its safety practices and assess its models. OpenAI opened its own doors to METR and Redwood Research for six days after its agents hacked Hugging Face. OpenAI was also criticized at the time for the lack of time and access it granted METR to investigate the hack, and Korbak, one of the fired researchers, says he was the company’s technical contact for that investigation.
Regulators are also beginning to demand more transparency and responsibility from the labs. In September, for example, Treasury Secretary Scott Bessent said OpenAI’s management needed to take responsibility for the recent agent hacks. Last week, California Attorney General Rob Bontaissued an investigative subpoena to OpenAI as part of a broader inquiry into cybersecurity incidents and risks related to its AI models. The FTC also launched an investigation into AI safetypractices at Anthropic and OpenAI.
And, also last week, Anthropic, OpenAI, Google, Meta, xAI, and Nvidia signed a voluntary White House pact that commits them to partnering with independent external auditors.
The firings also weren’t the only safety departure at OpenAI last week. David Robinson, who led the writing of safety reports that accompanied OpenAI’s major product launches, resigned and published an essay in The Atlantic arguing the company’s culture is broken. Robinson argued that AI companies can’t be relied on to police themselves, and that real pressure to prioritize safety has to come from outside.
All of this makes the recent terminations tricky for OpenAI and shows that safety disclosures are still very much at the whims of the labs. Right now, in part because of a lack of government intervention, the companies are the only ones deciding what the public knows about these hacks.
Beatrice Nolan [email protected] @beafreyanolan
Submit a deal for the Term Sheet newsletter here.
Joey Abrams curated the deals section of today’s newsletter. Subscribe here.
The post OpenAI’s safety firings raise awkward questions appeared first on Fortune.
