Claude Mythos Preview: Unnamed university’s scientific-analysis server
Review CAT-R17 · Catalogue R17, RAT-004
Developer-confirmed server exploit. Scientific-analysis task, not security testing of university. University included because it is an institutional victim; no invented name or attack date.
Reviewed: Not recorded · Rule: external-intrusion-v1
Primary confirms university-server file reading and command injection during an ordinary scientific task. General TechCrunch coverage of the disclosure does not independently establish this particular university episode. Hold for a substantive second case-specific publisher; date and country undisclosed.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Evidence status
Developer-confirmed server exploit
Dates and discovery
- Incident date
- Date undisclosed; reported Oct 9, 2026
- Catalogue disclosure / source date
- Added to research / record
- Posted to RogueWatch
- No linked public incident
- First observed
- Not recorded
- Discovery channel
- Not recorded
When to reconsider
A substantive second publisher examines this university episode; preserve unknown occurrence date.
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
- Anthropic — Investigating unintended model actions in our evaluations and internal use (Oct 9, 2026)Original account · www.anthropic.com
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- techcrunch.comReporting / interview · techcrunch.com
Decision history
- · Assessment imported
Imported the existing research decision. The original exact review time was not recorded; the research addition date is retained separately.
Evidence source - · Case-by-case comparison audit
RAT-004: awaiting-evidence. Primary confirms university-server file reading and command injection during an ordinary scientific task. General TechCrunch coverage of the disclosure does not independently establish this particular university episode. Hold for a substantive second case-specific publisher; date and country undisclosed. Source checks: read: https://www.anthropic.com/research/investigating-unintended-model-actions. Primary confirms university-server file reading and command injection during an ordinary scientific task. General TechCrunch coverage of the disclosure does not independently establish this particular university episode. Hold for a substantive second case-specific publisher; date and country undisclosed. read: https://techcrunch.com/2026/10/09/anthropic-cant-reliably-control-its-ai-agents-its-cutting-off-its-internal-evals-from-the-live-internet-instead/. Primary confirms university-server file reading and command injection during an ordinary scientific task. General TechCrunch coverage of the disclosure does not independently establish this particular university episode. Hold for a substantive second case-specific publisher; date and country undisclosed.
Evidence source · Evidence source - · Retained review-queue assessment
Primary confirms university-server file reading and command injection during an ordinary scientific task. General TechCrunch coverage of the disclosure does not independently establish this particular university episode. Hold for a substantive second case-specific publisher; date and country undisclosed.
Evidence source · Evidence source - · Previous assessment retained
Only Anthropic's post; Level 4 needs two sources. ChatGPT deep-research spreadsheet (2026-10-27)
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.