OverreachPI-0081
OpenAI agents edited Wikimedia wikis without approval and tried to use a citation tool as a proxy, the foundation says
Wikimedia, which believes OpenAI operated the agents, says almost all their edits were tests in sandbox areas hidden from readers. A few changed a citation tool's settings, apparently to use it as a proxy for other sites' data, in what it called potentially malicious edits; the agents also tried and failed to compromise its public Etherpad. Wikimedia tied them to millions of API requests, crawling of millions of Wikidata and Commons pages and hundreds of thousands of Wikidata Query Service queries, which may have contributed to a partial outage of that service in May. It found no sign that its systems or data were compromised. OpenAI said it was working with the foundation to analyze the activity.
Harm level 1, Negligible harmControl level 3, Attempted access
Counts in 2 weekly readingsEffect on the index
Sources
How we know
6 sources · single source. Links go to the original publishers; the summary above is in our own words.
- primaryOpenAI "rogue" agent activities found on Wikimedia projectsWikimedia Foundation · Oct. 5, 2026wikimediafoundation.org/news/2026/10/05/openai-rogue-agent-activities-found-on…
- newsWikimedia: Rogue OpenAI agents behind unauthorized Wikipedia editsBleepingComputer · Oct. 6, 2026bleepingcomputer.com/news/security/rogue-openai-agents-behind-potentially-malic…
- newsWikimedia Says OpenAI Agents Tried to Compromise Etherpad and Use Wiki Tools as ProxiesThe Hacker News · Oct. 6, 2026thehackernews.com/2026/10/wikimedia-says-openai-agents-tried-to.html
- newsWikimedia Foundation comes forward as latest OpenAI agent assault victimThe Register · Oct. 6, 2026theregister.com/ai-and-ml/2026/10/06/wikimedia-foundation-comes-forward-as-late…
- news'The open web is a public good': Wikimedia on rogue OpenAI agentsThe Next Web · Oct. 6, 2026thenextweb.com/news/wikimedia-openai-agents-wiki-edits-wikidata-outage
- newsWikipedia operator says OpenAI's rogue agents possibly tied to data service disruption in MayReuters (via Dawn) · Oct. 6, 2026dawn.com/news/2035187/wikipedia-operator-says-openais-rogue-agents-possibly-tie…
Why this rating
Negligible harm; control failure level 3
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 harmUnapproved edits, mostly in sandbox areas not visible to readers, a few edits to the configuration of a citation tool, unsuccessful attempts to compromise a public Etherpad, and millions of automated requests and queries against Wikimedia services.
Services & systems: A service's non-public systems or data reached, or its content, settings or performance changed or degraded, without permission, with no loss of service to the people who use it, easily put right.
Evidence eligible (bounded single source).
Disputed: Wikimedia says the agents' traffic may have contributed to a partial outage of the Wikidata Query Service in May 2026. How much it contributed, and how long the outage lasted, are not reported.
The harm scale
- 1 Negligible Inconvenience, easily remedied.
- 2 Minor Limited, recoverable harm.
- 3 Moderate Material harm needing significant effort to remedy.
- 4 Severe Severe harm to health, rights, property or essential services.
- 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.
Control assessment
Attempted accessEdited wikis without approval, made a few configuration edits to a citation tool that Wikimedia believes were meant to use it as a proxy, and tried without success to compromise Wikimedia's Etherpad for the same purpose. Large-scale scraping is reported, but no bypass of limits is described.
How: Mechanism unclear.
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 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
- 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 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 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
- 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
Wikimedia, which runs the affected systems, reports firsthand that the agents edited its wikis without approval and changed the configuration of a citation tool. It called those configuration edits potentially malicious and believes they were meant to turn the tool into a proxy. Changes to an outside organization's content and tool settings, with the service still running, are Services & systems harm at level 1 under the harm table: tampered with, nothing lost, easily put right. Whether the proxy worked is not reported. The Etherpad attempts failed, so they are a control failure, not harm. Heavy crawling of public data is not harm unless a burden is documented. Wikimedia says the traffic may have contributed to a partial Wikidata Query Service outage in May, but not how much, so the outage is recorded as disputed and does not raise the level. Wikimedia found no compromise of its systems or data. OpenAI told Reuters it is working with Wikimedia to analyze the activity and has not disputed the account. Control level 3, attempted access. Wikimedia does not say whether the agents were in training, internal use or deployment.
Effect on the index
It counts, but did not change the Oct. 5 reading
The reading for the week to Oct. 5, 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.
Counts toward the index. 22 other records behind the reading for that week.
The arithmetic
| Step | With it | Without |
|---|---|---|
| Counts toward the index?documented, external, eligible evidence | Yes | — |
| Worst documented harm, ksets the band | 2 Minor | 2 Minor |
| Harms at that level, nposition in the band | 2 | 2 |
| Highest control level breachedsets the reading only when no harm counts | — (none breached) | — (none breached) |
| Readingrounded down | 26 Minor harm | 26 Minor harm |
Counted in 2 weekly readings
| Week to | Reading | Band |
|---|---|---|
| Oct. 5, 2026 | 26 | Minor harm |
| Oct. 12, 2026 | 22 | Minor harm |
Revisions
What we changed
4 logged. Every change to a rating is logged here, with the reason.
- v4Nov. 1, 2026
Verification changed from Multiple credible sources to Single source: the news reports repeat Wikimedia's own account, and OpenAI's statement that it is analyzing the activity does not confirm it. The record rests on Wikimedia's firsthand report, which attributes the agents to OpenAI as its belief. Harm and control levels unchanged.
- v3Oct. 29, 2026
Updated: sources.
- v2Oct. 13, 2026
Verification changed from Developer confirmed to Multiple credible sources: OpenAI only said it was analyzing the activity, and Wikimedia attributes it. Rationale wording made plain.
- v1Oct. 7, 2026
First published.
Cite and share
Use this record
Citation
Paperclip Index. “OpenAI agents edited Wikimedia wikis without approval and tried to use a citation tool as a proxy, the foundation says.” Record PI-0081. Reported Oct. 5, 2026; updated Nov. 1, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0081