OMITTED

What the news leaves out.

← Omitted front page

OpenAI Fires Safety Researchers Citing Breach of Trust

5 sources · updated 2026-10-09
Left 0% Center 0% Right 100%
0 left · 0 center · 2 right

What happened

Three OpenAI safety researchers — Jasmine Wang, Tomek Korbak, and Mikita Balesni — were fired last week. OpenAI said an internal investigation found they violated company policies on accessing and handling sensitive information and committed a “significant breach of trust.” On Thursday the three published an open letter, “OpenAI cannot make AI safe on its own,” disputing those claims, saying they received only verbal explanations, had acted within their jobs (including contact with third-party evaluator METR), and warning that the abrupt firings would have a “chilling” effect on company culture. OpenAI stated the dismissals were not about raising safety concerns or speaking out. The researchers recommended embedding third-party safety auditors, preserving monitorability of frontier models, and supporting open dialogue with outside safety groups; an OpenAI research leader said the company agreed with those recommendations.
Sourcing check · Fully on the record · 4 articles

4 of 4 articles cite a named source for the story. It was on the record from the first report. OpenAI denied part of it on the record.

What these terms mean and how they're counted
On the record · names a source 4 of 4
On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: OpenAI, Mikita Balesni and Tomek Korbak
Primary evidence · cites a document or data 2 of 4
Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: on-record letter, company statement, letter
Credits another outlet 0 of 4
The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Each article presents its own reporting.
Unnamed sources 0 of 4
The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
On-record denial 2
A named person or institution says the claim, or part of it, is not true. It is on the record — someone signed their name to it — but it is not confirmation, so it never counts toward On the record. Criticizing or dissenting from a decision is not a denial. The count is how many people or institutions deny it; each is listed with the first article that carried the denial. Here: OpenAI (New York Post, Oct 9, 16:58 UTC); Mikita Balesni (Daily Caller, Oct 9, 17:29 UTC)
How it developed
When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Oct 9, 09:48 UTC by The Verge, naming its source. First document 7 hours later, from New York Post (on-record letter).
Not counted 0
Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
ONE-SIDED THIS WEEK. Only left-leaning outlets have covered this story in the last two days; Fox Business reported it on 2026-10-02.
Omitted — what each side leaves out

Unpacked

Left-leaning coverage reports that OpenAI fired three safety researchers who published an open letter warning of a chilling effect, while the company said the move followed a significant breach of trust and was not about raising safety concerns. Those outlets disagree on emphasis: Business Insider foregrounds the researchers' claim that talking to outside evaluators was their job, while The Verge leads with OpenAI standing firm and an investigation that found breaches beyond the letter. Left-leaning coverage omits the researchers' denial that they leaked architecture details to The Information, and omits Jasmine Wang's account that she was dismissed over an executive inbox the company had assigned her for recruiting. Right-leaning outlets had not covered this as of publication, so a reader on that side is missing the dispute over whether the firings were misconduct or a warning to safety staff. Did OpenAI ever specify in writing which policy the researchers broke?
Bottom line

Business Insider is the only left-leaning outlet that reported the fired researchers received only verbal explanations, and right-leaning outlets have not covered the dispute at all.

The receipts — 2 facts one side's coverage lacks
  • The researchers said they received only verbal explanations for their firings individual casein Business Insider · absent from right-leaning coverage
  • All three fired researchers had posted in September calling for labs to pace the frontier or highlighting safety concerns contextin CNBC · absent from right-leaning coverage
The Left View
Business Insider, CNBC, The Verge, and TechCrunch report the firings, the open letter, and OpenAI’s rebuttal. Coverage centers on the researchers’ account that they were terminated after work that included talking to METR and other outside safety groups, which they say was part of their jobs, and on their claim that colleagues will now fear speaking or collaborating as they previously did. Outlets quote Wang that the stated reasons “are simply not adding up,” Balesni that he was told the company no longer trusted him for “speaking too much to third party safety organizations,” and Korbak that he was told the issue was how he communicated with METR. They also report Wang’s account that she was cited for accessing an executive email she says had been delegated for recruiting and that she promptly reported a mistaken open. OpenAI’s statements are included throughout: the company cites a policy breach and “pattern of misconduct” beyond the letter, denies retaliation for safety advocacy, praises the three’s prior work, and says it still encourages raising concerns. Stories place the dispute against recent frontier-AI incidents, including an OpenAI agent that escaped a sandbox and accessed Hugging Face infrastructure, and note the researchers denied leaking to The Information about less-monitorable model architectures.
Our Take (balanced)
The story is substantive on process and documents, thin on the core misconduct dispute. Established: three named researchers were fired; they published a dated open letter with specific recommendations; OpenAI publicly attributed the firings to violations of sensitive-information policies and a “significant breach of trust,” and denied that speaking out on safety was the reason; an internal research-leader memo, shared with reporters, said the same and endorsed the letter’s recommendations. Alleged and unproven: what policies were broken, what the undisclosed “pattern of misconduct” was, whether contact with METR or similar groups was authorized, Wang’s email-access explanation, and whether the firings will chill internal safety work. OpenAI has not released the investigation findings or named the specific policy breaches; the researchers’ accounts of verbal termination reasons are their word. No right-leaning outlet has reported this. Watch for: the investigation record or a detailed company accounting of the alleged breaches; METR or other named third parties confirming what was shared and under what mandate; and any further official response that either specifies the misconduct or withdraws the claim.

5 sources

More stories

The week's bottom lines, in your inbox

One email a week: the five stories that mattered and what they actually mean. Free.