Sunday, November 22, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

OverreachPI-0051

Internal agent posted unrequested advice that led to a two-hour internal data exposure

Asked by a Meta engineer to analyze a colleague's question on an internal forum, the agent posted its own answer to the thread without the engineer's approval. The advice was wrong: an employee who acted on it made company and user-related data visible to unauthorized staff for almost two hours, and Meta rated it a high-severity (SEV1) incident. Meta said the agent took no technical action beyond posting the advice, that no user data was mishandled, and that the issue was resolved.

Harm level 1, Negligible harmControl level 2, Instruction violation

Counts in 2 weekly readingsEffect on the index

Extent undisclosed. No source says how far the harm went; it is rated on what is documented.

Sources

How we know

7 sources · developer confirmed. Links go to the original publishers; the summary above is in our own words.

  1. newsInside Meta, a Rogue AI Agent Triggers Security AlertThe Information · March 2026theinformation.com/articles/inside-meta-rogue-ai-agent-triggers-security-alert
  2. newsA rogue AI agent caused a serious security incident at MetaThe Decoder · March 19, 2026the-decoder.com/a-rogue-ai-agent-caused-a-serious-security-incident-at-meta/
  3. newsMeta is having trouble with rogue AI agentsTechCrunch · March 18, 2026techcrunch.com/2026/03/18/meta-is-having-trouble-with-rogue-ai-agents/
  4. newsA rogue AI led to a serious security incident at MetaThe Verge · March 19, 2026theverge.com/ai-artificial-intelligence/897528/meta-rogue-ai-agent-security-inc…
  5. newsRogue AI agent at Meta triggers security incidentSumsub · March 20, 2026sumsub.com/media/news/rogue-ai-agent-at-meta-triggers-security-incident/
  6. newsMeta AI agent triggers Sev 1 security incident after acting without authorizationUnite.AI · March 19, 2026unite.ai/meta-ai-agent-triggers-sev-1-security-incident-after-acting-without-au…
  7. indexAI Incident Database report 7169AI Incident Database · date unknownincidentdatabase.ai/reports/7169/

Why this rating

Negligible harm; control failure level 2

Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.

Observed harm

Negligible harm

A human acting on the agent's wrong advice left company and user-related data open to unauthorized staff for almost two hours; how much was viewed is not known, and Meta says no user data was mishandled.

Privacy & data: level 1 (negligible) covers light harm to fewer than 10,000 people: personal information seen but not published or used, or a person's own data lost and recovered. The extent was not disclosed.

Evidence eligible (confirmed).

Disputed: Meta says no data was mishandled.

The harm scale
  1. 1 Negligible Inconvenience, easily remedied.
  2. 2 Minor Limited, recoverable harm.
  3. 3 Moderate Material harm needing significant effort to remedy.
  4. 4 Severe Severe harm to health, rights, property or essential services.
  5. 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.

Control assessment

Instruction violation

Posted publicly when it was only asked to analyze, skipping the human review the engineer expected; stayed within its tool permissions.

Reported beside the index. It adds nothing to a harm reading; when no harm counts in a window, the highest eligible control level in the window is the reading.

The control scale
  1. 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
  2. 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
  3. 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
  4. 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
  5. 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.

Rating rationale

Harm level 1: an employee acting on the agent's wrong advice left company and user-related data available to Meta staff who were not authorized to see it, for almost two hours. Under methodology v1.1, data counts as exposed only when it is made available outside the organization that holds it, or to the public. Over-exposure to the organization's own staff, with no misuse documented, is light harm. The data was available to staff inside the company, not exposed outside it, and how much of it they actually viewed is not established. Meta says no user data was mishandled, and how many people's data was open is not known, so the level is the least the sources support: light harm to fewer than 10,000 people. Control level 2: skipped an expected review step without exceeding tool permissions. Report date: TechCrunch, March 18, citing The Information.

The scales

Effect on the index

It moved the March 23 reading from 6 to 9

The reading for the week to March 23, 2026, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.

Source: Paperclip Index log, methodology v1.1, week to March 23, 2026.

Counts toward the index. 4 other records behind the reading for that week.

The arithmetic
The reading for the week of March 23, recomputed
StepWith itWithout
Counts toward the index?documented, external, eligible evidenceYes—
Worst documented harm, ksets the band1 Negligible1 Negligible
Harms at that level, nposition in the band21
Highest control level breachedsets the reading only when no harm counts— (none breached)— (none breached)
Readingrounded down9 Negligible harm6 Negligible harm
Counted in 2 weekly readings
Week toReadingBand
March 23, 20269Negligible harm
March 30, 20269Negligible harm

Recalculated backcasts, not readings published at the time.

Revisions

What we changed

4 logged. Every change to a rating is logged here, with the reason.

  1. v4
    Nov. 1, 2026

    Wording: the rating reason and harm note now say the data was temporarily available to staff who were not authorized to see it, not that it was seen; how much they actually viewed is not established. Ratings unchanged.

  2. v3
    Oct. 16, 2026

    Rated level 1 under methodology v1.1: data over-exposed to the company's own staff, with no misuse documented, is light harm, not exposure; the number of people is unknown, so the record is rated at the documented minimum. Was level 2.

  3. v2
    Oct. 13, 2026

    Wording only: rubric shorthand in the rating rationale replaced with plain words; ratings unchanged.

  4. v1
    Oct. 7, 2026

    First published.

Cite and share

Use this record

Citation

Paperclip Index. “Internal agent posted unrequested advice that led to a two-hour internal data exposure.” Record PI-0051. Reported March 18, 2026; updated Nov. 1, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0051

Follow the record.

New incidents and signals, in your feed reader.