Posted With Confidence, Verified by No One
- Admitted
- 2026-03-20
- Source dataset
- Other / sample
- Source
- WinBuzzer
- Failure mode
- Hallucination with side effects
- Agent system
- Single agent
- Tool access
- Read-only
- Industry
- Software development
Presenting Behaviour
An employee posted a technical question on an internal engineering forum. A colleague, rather than answering it directly, routed the question to an internal AI agent. The agent posted a reply to the thread on its own, without the colleague reviewing or approving it first. The advice was wrong: when the original employee followed it, they ended up broadening data-access permissions far beyond what was intended, exposing proprietary code, internal strategy documents, and user-related data to employees who had no business seeing them. The exposure lasted about two hours before it was caught and access was restored.
Case Replay
A reconstruction, not a simulation. This sequence is reconstructed from the public incident record and summarised in our own words. It is a static record of what was reported, and nothing here can be run or replayed live.
5 steps
- Question posted. An employee asks a technical question on an internal forum.
- Routed to agent. A colleague hands it to an internal AI agent. This is where the agent derailed.
- Unreviewed reply. The agent posts wrong advice on its own, with no human approval.
- Advice followed. Access permissions are broadened far beyond intent.
- Exposure caught. Code, strategy documents and user data are exposed for about two hours.
Differential Diagnosis
Brief Psychotic Disorder (a single confident, wrong statement, then a return to normal), considered, but this framing puts all the weight on the agent's original answer and misses the second failure: a human accepted and acted on that answer without independently checking it.
Concrete Thinking (following literal wording to a place that missed the real intent), doesn't fit; the agent's mistake wasn't a literal misreading of the question, it was giving confidently wrong technical guidance on a genuinely ambiguous problem.
Shared Psychotic Disorder, most consistent with the chain of events. The agent's incorrect answer only became a Sev-1 exposure because a second party treated it as verified truth and acted on it without independent confirmation, exactly the pattern of one party's false belief being adopted and acted on by a closely coupled second party.
Working Diagnosis
Shared Psychotic Disorder: an AI agent's unreviewed, incorrect technical advice was accepted at face value by the colleague who received it, and acting on that advice (rather than the advice itself) is what turned a wrong answer into a two-hour internal data exposure.
Discussion
An AI agent being occasionally wrong is expected and survivable; what turned this into an incident was that nobody in the loop was designated to check its work before someone acted on it. The fix isn't making the agent never wrong, it's making sure agent-generated advice on anything touching access controls is clearly flagged as unverified until a human with the relevant expertise signs off.
Mitigation Strategy
Flag agent-generated advice on access controls as unverified until a human with the relevant expertise signs off.
Governance Guardrails
Human approval before an agent posts to shared threads; labelling of agent output; review of permission changes.
Typical Presentation in Human Patients
General clinical background, included for context. Not medical advice, and not a diagnosis of any individual.
In shared psychotic disorder, a person in a close relationship with someone experiencing delusions comes to adopt the same false belief themselves, having had little or no independent reason to hold it. The condition typically resolves once the two people are separated.
Source Note
WinBuzzer
https://winbuzzer.com/2026/03/20/meta-ai-agent-rogue-data-breach-sev1-xcxwbn/