OpenAI fires three safety researchers over outside info sharing
A firing at OpenAI set off a whistleblowing fight on Thursday, SoftBank's money into the frontier lab is fully in, and a court handed Google's AI Overviews a win over the publishers it cannibalizes.
OpenAI has parted ways with three researchers on its safety team for allegedly sharing confidential company information with an outside AI safety organization, the Wall Street Journal reported — a move that drew an immediate reaction from Congressman Greg Casar, who said it "looks like they're firing whistleblowers." OpenAI's spokesperson said the company "parted ways with three individuals for violating our policies on accessing and handling sensitive company information," and that an internal investigation found they "mishandled sensitive information outside established company procedures." The Journal report does not name the researchers, the outside organization, or what information was shared; posts on X identified several OpenAI departures in the safety and chain-of-thought-monitoring group, but TechCrunch has not confirmed their identities, so treat the names as unconfirmed. What makes this more than an HR story is the timing: it lands two days after the New York Times reported that executives had brushed aside employees' security warnings, in the middle of a run of rogue-agent incidents that pushed OpenAI to scrap GPT-6.1 Astra, and against precedent — OpenAI fired Leopold Aschenbrenner and Pavel Izmailov for alleged leaks in 2024, and Aschenbrenner has called that pretextual. If these researchers went outside because internal channels failed them, this is a governance story; if they genuinely leaked, OpenAI still has to prove it, because right now the company's word is the only evidence anyone outside has. Casar says he is sending OpenAI a demand for transparency, which means Congress intends to find out which story is true.
SoftBank has completed the third and final $10 billion tranche of its $30 billion pledge to OpenAI's last funding round — and The Information reports Nvidia has now made its own final $10 billion payment. SoftBank confirmed the tranche in its own press release: cumulative investment in OpenAI reaches $64.6 billion for roughly 13 percent of the company, funded largely by a $11.1 billion high-yield bond sale last month — the largest such deal globally. The Nvidia half is single-sourced so far, reported by The Information citing a person familiar with the investments, with no second outlet or company confirmation; Nvidia does not disclose its tranches publicly, so we are taking it at report's value. The round itself was announced in February at up to $122 billion on an $852 billion valuation, anchored by Amazon, Nvidia and SoftBank, so what today's tranche confirms is that the record round closed fully funded. We covered how SoftBank was paying for it in SoftBank is funding OpenAI with junk.
A US judge dismissed the Chegg and Penske Media antitrust lawsuits alleging Google forced them to accept AI Overviews in exchange for search indexing, reducing their web traffic. The ruling, reported by Reuters and covered by The Verge, throws out publishers' argument that conditioning inclusion in Google's index on tolerating AI summaries of your content is an antitrust violation. Chegg had argued the summaries answered the homework questions its site existed to answer; Penske, which publishes Rolling Stone, said traffic to its sites was being siphoned off. Dismissal at the trial level is not necessarily the end — appeals remain possible — but it is the second straight signal that the "AI Overviews are extortion" theory is struggling in court even as Google signs traffic deals with publishers on the side. Between this and Google paying roughly 100 publishers for AI answers without disclosing the price, the market for AI-borrowed content is quietly splitting into the ones with contracts and the ones with lawsuits, and today the lawsuits just got weaker.
What to watch: whether OpenAI names what was actually shared — and whether the three researchers speak first.
When a lab fires safety staff for talking to outside organizations, should the burden of proof sit with the company? Tell us in the comments.
