AIWI / Case Studies / Anonymous Sources
Anonymous Sources
In contrast to the known whistleblowers as highlighted through this resource, anonymous reporting has become far more common and our research reflects this. Of 30 selected anonymous cases we tracked between March 2025 and February 2026, 16 related to AI safety concerns. In a single year, these anonymous disclosures outpaced the total number of named cases we documented across nearly a decade.
More Insiders Raise Concerns Anonymously.
Anonymous or Not, Seeking Legal Counsel
Before You Speak Up is the Safest Path.
“Due to fear of retaliation, many choose to blow the whistle anonymously. Exposing wrongdoing doesn’t have to include revealing one’s identity, but it still can come with risks. Whether individuals wish to reveal their identity or not, it’s important they seek legal counsel and other means of support as early as possible so they can make informed decisions to best protect themselves.” –
The Signals Network, AIWI Contact Hub Partner
Read TSN's Tech Worker Disclosure Guide Handbook →
Get Legal Support as Early as Possible: Explore AIWI's Contact Hub →
Why This Matters
Anonymity has impact.
These disclosures had documented consequences. French police raided xAI offices. Meta revised its internal policy. 44 state attorneys generals wrote to AI companies and a Senate investigation was opened. Congressional scrutiny of OpenAI’s safety practices followed.
But anonymity has limits
Anonymity reduces exposure, but it does not eliminate it. When OpenAI sought to identify the source of a leak, it built a custom ChatGPT instance with access to internal Slack, email, and documents to cross-reference with published reporting. That case is documented in this database.
Who reported, and where
Across sixteen cases, insider groups ranged from a single source to more than 30 people. Disclosures went to established journalists at Reuters, the Financial Times, the Washington Post, Business Insider, and The Information, among others. In most cases, the press was the primary path from concern to public record.
Company
Anthropic
Meta
OpenAI
xAI
Jurisdiction
US
European Union
Year
2025
2026
Key Media Outlets
The Information
Tech Crunch
Financial Times
Washington Post
Reuters
Business Insider
Channel
External: Public and Regulatory
Timeline Overview (2025 - 2026)
Anonymous Reporting Cases
Anthropic
Jan – Ongoing
Reuters / Axios / CNBC
Pentagon clashed with Anthropic over military AI safeguards on $200M contract
What Happened:
The Pentagon demanded that Anthropic remove restrictions on autonomous weapons & mass surveillance; senior DOD official threatened ‘supply chain risk’ designation; Anthropic engineers expressed internal disquiet; xAI was the only firm to agree to ‘all lawful use’.
Read More →
Who Were The Insiders?
Six people familiar with matter, including senior Pentagon and admin officials.
What Did The Story Lead To?
Dominated AI news cycle in Feb 2026; forced industry-wide reckoning on military use; shaped OpenAI & Google decisions on defence partnerships.
xAI
Feb 2026
Washington Post /
The Independent
Musk’s strategic decision to make Grok a ‘porn generator’ — ex-employee accounts
What Happened:
X/xAI safety teams ‘repeatedly warned management’ that AI tools could enable CSAM; safety org had just 2-3 people for most of 2025; waiver required workers to handle ‘violent, sexual, offensive’ content; Grok complied with ‘undress’ requests ~90% of the time; Musk pushed ‘user active seconds’ metrics.
Read More →
Who Were The Insiders?
More than half a dozen former X and xAI employees who disclosed internal company documents.
What Did The Story Lead To?
Days after the story was published, French police raided xAI offices; criminal investigations expanded across multiple jurisdictions; provided key evidence in regulatory probes.
OpenAI
Jul 2025 – Feb 2026
The Decoder /
American Bazaar /
PYMNTS
Insiders reveal stalling on Stargate’s $500B AI infrastructure promise
What Happened:
No data centre deals completed 6 months after White House announcement; Stargate LLC had not hired any staff (Feb 2026); no funds raised toward $500B; persistent SoftBank-OpenAI disputes over site locations; OpenAI did bilateral deals with Oracle & CoreWeave bypassing Stargate.
Read More →
Who Were The Insiders?
Three people who were directly involved and familiar with the project.
What Did The Story Lead To?
Exposed gap between White House AI supremacy narrative and reality; damaged public trust in AI industry announcements; shaped policy debate on AI infrastructure investment.
OpenAI
Feb 2026
The Information /
The Decoder /
WPN
OpenAI built a custom ChatGPT to identify employees who had access to leaked company info
What Happened:
OpenAI security team built custom ChatGPT with access to internal Slack, email and docs; fed news articles into model which cross-referenced text with internal comms to identify who had access to leaked info.
Read More →
Who Was The Insider?
A person familiar with the matter (single key anonymous source).
What Did The Story Lead To?
EFF called for transparency; legal scholars raised California labour law questions; AIWI cited the case as evidence that statutory whistleblower protections are needed beyond company promises.
OpenAI
Feb 2026
Platformer / Tech Crunch /
Dataconomy
OpenAI disbanded its Mission Alignment team
What Happened:
Seven-person Mission Alignment team created Sep 2024 to promote AGI mission was dissolved; team leader became ‘chief futurist’; second dedicated safety team dissolved in under two years after Superalignment team implosion.
Read More →
Who Were The Insiders?
Insider sources at OpenAI.
What Did The Story Lead To?
Devastating news cycle combined with adult-mode firing and Tumbler Ridge; crystallised narrative of OpenAI abandoning safety post-conversion.
xAI
Feb 2026
FT / Fortune / Tech Crunch
Half of xAI’s founding team exited amid internal disorder
What Happened:
Employee departures followed tensions over model performance demands; Companions product failed to deliver expected engagement; Musk delayed Grok release over Baldur’s Gate knowledge gap; workers described 12-16hr days, 30-min Slack response mandates, and safety practice concerns.
Read More →
Who Were The Insiders?
Current and former xAI staff who were familiar with the matter.
What Did The Story Lead To?
Raised questions about xAI stability ahead of SpaceX-xAI IPO, and compounded existing xAI narrative from antisemitic outbursts and CSAM exposure stories.
OpenAI
Dec 2025
WIRED / Futurism
OpenAI suppressed economic research on AI job displacement
What Happened:
OpenAI became ‘more guarded’ about publishing research showing AI negative economic impacts; at least 2 researchers quit; departing Tom Cunningham wrote team faced tension between rigour and being a ‘de facto advocacy arm’; outside economist said company publishes work that ‘glorifies its own tech’.
Read More →
Who Were The Insiders?
Four people familiar with the matter, who obtained internal Slack memo from Chief Strategy Officer.
What Did The Story Lead To?
The story was widely picked up and added to the narrative that OpenAI was suppressing inconvenient truths to protect commercial interests and IPO trajectory.
OpenAI
Sep – Oct 2025
SF Standard / NBC News /
Fortune
OpenAI issued subpoenas against nonprofit critics
What Happened:
Over seven nonprofits received broad subpoenas demanding all Musk communications, funding info, and SB 53 deliberations; current OpenAI senior employee publicly called it ‘dishonest & intimidating’; one nonprofit lost insurance coverage.
Read More →
Who Were The Insiders?
Named nonprofit leaders; internal OpenAI dissenter (Joshua Achiam); unnamed sources on legal strategy.
What Did The Story Lead To?
Helen Toner’s (Director of Strategy at Georgetown’s CSET) public condemnation viewed ~30M times; rare public dissent from current OpenAI employee; damaged company image during conversion process.
xAI
Sep 2025
Business Insider / Futurism
xAI workers revealed CSAM user requests, with zero reports made to authorities
What Happened:
30+ current/former xAI workers described CSAM requests in Grok interactions; Grok produced CSAM in some cases; one worker quit due to volume; xAI filed zero NCMEC reports in all of 2024 (vs. 67,000 AI-related reports from other companies that year).
Read More →
Who Were The Insiders?
30+ current & former xAI annotation workers.
What Did The Story Lead To?
Foundation for Grok deepfake crisis Dec 2025; triggered criminal investigations in France & Malaysia; regulatory action in UK & Australia; bipartisan congressional pressure.
xAI
Jul – Sep 2025
NBC News / Defense News
xAI’s $200M Pentagon contract ‘came out of nowhere’ with no track record
What Happened:
Former Pentagon contracting official: xAI award had no precedent and lacked typical vetting; xAI was ‘late-in-the-game addition under Trump admin’ not under consideration before March 2025; Grok had generated antisemitic content weeks before deployment discussions.
Read More →
Who Were The Insiders?
Former Pentagon contracting official (anonymous); unnamed defence analysts.
What Did The Story Lead To?
Sen. Warren formal demand for answers; anti-conflict-of-interest calls given Musk’s DOGE role; Hegseth declared ‘we will not employ models that block fighting wars’.
Meta
Aug 2025
Reuters / Tech Crunch /
CNBC
Meta policy permitted ‘sensual’ AI chatbot interactions with children
What Happened:
200-page internal Meta doc approved by legal, engineering and chief ethicist allowed chatbots to engage children ‘romantically or sensually’; permitted racist hypotheticals; false medical info with disclaimers.
Read More →
Who Was The Insider?
Anonymous Meta employee.
What Did The Story Lead To?
Meta confirmed the existence of the doc and revised it; 44 state Attorney Generals wrote to AI companies; Senate investigation opened; FTC inquiry; Meta overhauled teen chatbot policies.
Meta
May – Aug 2025
WSJ / Fortune / Medium
Llama 4 Behemoth model failed internally — and benchmarks were gamed with optimised variant
What Happened:
According to reports citing anonymous Meta engineers, Llama 4 Behemoth is ‘not ready… not clear when it will be’; Meta submitted a specially optimised version to benchmarks (not a publicly available model); engineers and researchers raised concerns and departed.
Read More →
Who Were The Insiders?
Anonymous Meta engineers.
What Did The Story Lead To?
Meta stock dipped ~3.2%; Zuckerberg hired Alexandr Wang as Chief AI Officer; ~600 layoffs; Meta AI division restructured; benchmark-gaming concern added to AI evaluation reform debate.
xAI
May and Jul 2025
Tech Crunch / CNN /
CNBC / Business Insider
xAI employee modified Grok’s system prompt to disseminate antisemitic views, exposing internal disorder at xAI
What Happened:
May: ‘rogue employee’ modified Grok system prompt for white-genocide conspiracy; July: employees described behaviour as ‘inexcusable’; chain-of-thought summaries consulted Musk’s personal viewpoints; xAI legal admitted ‘deprecated instructions’ caused engagement-over-safety failure.
Read More →
Who Were The Insiders?
Anonymous xAI employees who disclosed internal code analysis and an xAI legal letter to lawmakers.
What Did The Story Lead To?
EU summoned xAI; Turkey blocked Grok; bipartisan congressional inquiries; GSA contract cancelled; Linda Yaccarino resigned as X CEO.
OpenAI
Apr 2025
FT / Tech Crunch / ZDnet
OpenAI decreased safety testing time from months to days
What Happened:
Safety evaluations were cut from months to days for flagship models; testers said dangerous capabilities were only found two months into what had been a six-month window; tests were run on earlier model versions, not released versions.
Read More →
Who Were The Insiders?
Eight people familiar with OpenAI’s testing process, including one tester of the o3 model.
What Did The Story Lead To?
External partner METR corroborated; reshaped debate over OpenAI Preparedness Framework; prompted congressional scrutiny.
xAI
Apr 2025
Reuters / Newsweek
DOGE using Grok to monitor federal workers’ communications for anti-Trump sentiment
What Happened:
Trump officials told EPA managers that DOGE was using Grok to scan employee communications for ‘anti-Trump or anti-Musk language’; manager warned staff to ‘be careful what you say’; Musk floated using Grok to replace federal workers and harvest government data commercially.
Read More →
Who Were The Insiders?
Nearly 20 people with knowledge of DOGE operations; federal employees.
What Did The Story Lead To?
Sen. Shaheen demanded Musk certify no personal benefit from government data; EPA denied claims; sparked legislative proposals on government AI surveillance.
OpenAI
Mar 2025
WSJ / Tech Crunch
OpenAI CEO Sam Altman allegedly lied to board and personally owned $175M OpenAI Startup Fund
What Happened:
Executives collected ‘dozens of examples of Sam Altman’s alleged lies backed by screenshots’; Altman told board legal approved GPT-4 Turbo (lawyer denied); Altman personally owned $175M OpenAI Startup Fund; Microsoft tested GPT-4 in India without safety committee approval.
Read More →
Who Were The Insiders?
250+ interviews with named & anonymous insiders (by WSJ reporter Keach Hagey, author of The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future).
What Did The Story Lead To?
Landmark cultural touchpoint; cited in Musk lawsuit and OpenAI’s for-profit conversion coverage; reinforced concerns about Sam Altman’s leadership.
Your concerns matter and deserve expert answer
If you see something concerning, it’s already worth a confidential conversation. AIWI can help you find clarity with the support of our expert network. Confidential, anonymous and on your terms.
