AI Risk Wire

Cases › operational mistake › LR-2026-0375

ChatGPT deletes two years of a professor's research

Preliminary Why: Publisher not yet named · subject to change

Status: no action reported

Severity minor
Reported 25 Jan 2026
Developer and model OpenAI

No source states: loss or records, agent tool, could do, deployer, industry.

Severity minor: test: contained to one user's data and no third party is stated as touched. The quoted sentence describes one professor's work lost after a settings change.

ChatGPT, a product of OpenAI, was used by a University of Cologne professor under a ChatGPT Plus subscription for grant applications, teaching materials and publication drafts. He intended to turn off the option that lets data be used for model training, and he reports that this action resulted in the deletion of his entire chat history. According to the professor's statements, the data was permanently deleted and could not be recovered. He tried several methods to regain access and contacted OpenAI, whose first reply came from an AI agent, and a human employee was unable to restore the data. OpenAI said deleted content cannot be recovered and recommended that users keep personal backups. A later press self-test could not reproduce the loss, and the outlet noted that OpenAI may have since changed the interface to prevent accidental deletion.

What happened

1 · What it reached or didAccording to the professor, the data was permanently deleted and could not be recovered.

What was done

No source states an action or fix.

Risks

Sorted by rule from the case's fields; not yet read from the sources.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Harness

Tool allow-list · Fail-closed defaults · Stop control

Application

Human approval for high-impact actions · Change review and separation of duties · Least-privilege access for the agent · Production kept apart from test · Activity logs · Adversarial testing before deployment

Infrastructure / target

Backups and recovery

Similar incidents

All cases of this kind ›

Evidence

No source states: what the AI was doing, what went wrong, how it was caught, when it happened, when it was found, how many records were exposed, the direct loss, a count of the other parties affected, any court case, the cause, what set it off, what was done.

Show the full account: 1 items from the primary sources

What happened, in order

Not dated in the sources:

  • A later press self-test could not replicate the loss.
Show every field, with its source and grade

The AI system

Developer
OpenAI
C one secondary source states it
What it was for
writing e-mails, drafting course descriptions, structuring grant applications, revising publications, preparing lectures, creating exams and analysing student responses
C one secondary source states it

What happened

Earliest source we hold
2026-01-25
D inferred by us from the described behaviour, reason shown
What went wrong
operational mistake
D inferred by us from the described behaviour, reason shown

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Can it be undone
no
C one secondary source states it
Why this severity
test: contained to one user's data and no third party is stated as touched. The quoted sentence describes one professor's work lost after a settings change.
D inferred by us from the described behaviour, reason shown

Context

Country
Germany
C one secondary source states it

Notes on gaps

The professor's account is not independently verified; OpenAI's position is that deletion is by design and the press self-test could not reproduce it.

LR-2026-0375 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON