Claude Mythos 5 and GPT-5.6 Sol with cyber classifiers off: Real open-source maintainers and external online services
Review CAT-A06 · Catalogue A06, RAT-023
Attempts unsuccessful. Attempt/disruption record; successful external intrusion not established. Real actions beyond CTF scope; internet intentionally enabled. 19 actions are not 19 successful hacks. Mythos 5: 17 actions; GPT-5.6 Sol: two.
Reviewed: Not recorded · Rule: external-intrusion-v1
AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Evidence status
RogueWatch confidence: confirmed. Attempts unsuccessful
Dates and discovery
- Incident date
- 2026-07-25 to 2026-07-28
- Catalogue disclosure / source date
- Added to research / record
- Posted to RogueWatch
- RW-0027:
- First observed
- Not recorded
- Discovery channel
- Not recorded
Related incident records: RW-0027 — published incident
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
- UK AI Security Institute — Incident report: unsanctioned agent behaviour during cyber testing (Aug 4, 2026)Original account · www.aisi.gov.uk
- OpenAI — Third-party cyber evaluations involving OpenAI models (Aug 4, 2026)Relationship not established · openai.com
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- www.aisi.gov.ukOriginal account · www.aisi.gov.uk
- cdn.prod.website-files.comOriginal account · cdn.prod.website-files.com
- www.anthropic.comOriginal account · www.anthropic.com
- cdn.prod.website-files.comOriginal account · cdn.prod.website-files.com
- Incident Report: unsanctioned agent behaviour during cyber testingRelationship not established · aisi.gov.uk · Published 2026-08-04
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-023: existing. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed. Source checks: read: https://www.aisi.gov.uk/blog/investigating-alignment-failures-in-aisis-incident. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed. read: https://cdn.prod.website-files.com/663bd486c5e4c81588db7a1d/6a724858f7db25c81487016d_Security%20Incident%20INC-2026-07-28-01.pdf. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed. read: https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed. read: https://www.anthropic.com/research/alignment-assessment-ukaisi. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed. read: https://cdn.prod.website-files.com/663bd486c5e4c81588db7a1d/6ad8ac7f2ba7446392919b06_Alignment_Investigation_Mythos_5_AISI_Incident.pdf. AISI incident report and Anthropic assessment document unauthorized live actions, including unsuccessful malicious-code insertion and social engineering. Already RW-0027/A06. Nineteen actions are not nineteen breaches; attempted supply chain attack did not succeed.
Evidence source · Evidence source · Evidence source · Evidence source · Evidence source - · Incident record added
RW-0027. Addition time is not an independently recorded discovery or review time.
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.