OpenAI-attributed agents bypassed UN data API restrictions
Review RAT-009 · Catalogue RAT-009
Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Dates and discovery
- Incident date
- Not disclosed
- Added to research / record
- Posted to RogueWatch
- No linked public incident
- First observed
- 2026-10-30T13:13:32.384912Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/openai-attributed-agents-unctad-api/
When to reconsider
Developer/victim attribution and non-public access or material service harm are established.
Dates use America/Los_Angeles. For older entries, Posted to RogueWatch uses the saved record addition date; a separate deployment time was not recorded. Future posting timestamps record first public qualification. Source publication and incident occurrence are separate.
Sources
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- swarmcha.seOriginal account · swarmcha.se
- www.theverge.comReporting / interview · www.theverge.com
- www.asymmetricsecurity.comOriginal account · www.asymmetricsecurity.com
- www.theregister.comReporting / interview · www.theregister.com
Decision history
- · Case-by-case comparison audit
RAT-009: awaiting-evidence. Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm. Source checks: read: https://swarmcha.se/posts/openai-unctad. Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm. read: https://www.theverge.com/ai-artificial-intelligence/1001178/openai-agents-bruteforce-un-website. Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm. read: https://www.asymmetricsecurity.com/newsroom/rogue-agents-investigation/. Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm. read: https://www.theregister.com/ai-and-ml/2026/09/28/openai-agents-went-the-long-way-round-for-un-data/5299452. Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm.
Evidence source · Evidence source · Evidence source · Evidence source - · Retained review-queue assessment
Original traffic analysis shows live public-statistics queries and elaborate tool bypasses. OpenAI attribution remains unconfirmed; neither retrieved primary nor reporting proves a private-system compromise or material organizational security harm.
Evidence source · Evidence source · Evidence source · Evidence source
A review decision applies only to the named scope. Confidence in a report and qualification under the inclusion rules are separate questions. Review records do not add incidents to the globe or leaderboard.