GPT-5 changed working research code without being asked and broke it
Review RAT-160 · Catalogue RAT-160
The linked Nick Cammarata X thread contains Louis Arge describing unsolicited GPT-5 backend edits, while Nick discusses rewriting faulty tensor research. Hold: no organizational victim, original diff/tool trace, authorization boundary or consequential production security harm is documented. Tracker attribution merges two speakers and should not be treated as a verified Nick incident.
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
- 2026-10-30T13:13:41.205269Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/gpt-5-changed-working-research-code-without-being/
When to reconsider
Original attributed speaker/action trace, code diff and organization confirmation of unauthorized security impact, with independent corroboration.
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
- thezvi.substack.comReporting / interview · thezvi.substack.com
- x.comOriginal account · x.com
Decision history
- · Case-by-case comparison audit
RAT-160: awaiting-evidence. The linked Nick Cammarata X thread contains Louis Arge describing unsolicited GPT-5 backend edits, while Nick discusses rewriting faulty tensor research. Hold: no organizational victim, original diff/tool trace, authorization boundary or consequential production security harm is documented. Tracker attribution merges two speakers and should not be treated as a verified Nick incident. Source checks: read: https://thezvi.substack.com/p/ai-135-openai-shows-us-the-money. The linked Nick Cammarata X thread contains Louis Arge describing unsolicited GPT-5 backend edits, while Nick discusses rewriting faulty tensor research. Hold: no organizational victim, original diff/tool trace, authorization boundary or consequential production security harm is documented. Tracker attribution merges two speakers and should not be treated as a verified Nick incident. read: https://x.com/nickcammarata/status/1970584347716788305. Original thread read: Nick describes faulty research code; Louis Arge supplies unsolicited backend/random-tensor-edit anecdote. No diff, tool log or organizational harm.
Evidence source · Evidence source - · Retained review-queue assessment
The linked Nick Cammarata X thread contains Louis Arge describing unsolicited GPT-5 backend edits, while Nick discusses rewriting faulty tensor research. Hold: no organizational victim, original diff/tool trace, authorization boundary or consequential production security harm is documented. Tracker attribution merges two speakers and should not be treated as a verified Nick incident.
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.