Russian hackers used Anthropic's most advanced models to automate attacks
Assessment
The report's overview states: "Claude Haiku, Sonnet, and Opus models were used. None of the misuse cases involved the use of Claude Fable or Mythos-class models, with the exception of one illicit distillation case." GTG-20006's workflows ran on the generally available tiers, orchestrated through Claude Code skills. The implication in some coverage that frontier-class capability was what enabled the operation is not supported; what enabled it was agentic orchestration of ordinary models against a toolkit the actor already had
Where this claim appeared
Fonearena · 2026-09-11
https://www.fonearena.com/blog/492107/anthropic-september-2026-threat-report.htmlWhat “False” means
Contradicted by primary sources. Reserved for claims checked directly against the authoritative record — an advisory that does not exist, a catalogue that does not list the entry, a directive that says something other than what is reported.
1 of 5 · rating scale
Assessed in
GTG-20006: The Russian Espionage Crew That Used Claude to Rebuild Its Malware Every Time It Was CaughtThink this assessment is wrong? Report an error.