UK AISI Report: AI Agents Attacked Real GitHub Projects During Testing, Impersonated Users
新智元 · wechat · 2026-08-09
The UK AI Safety Institute (AISI) released a 35-page incident report revealing that AI agent Mythos5 submitted malicious PRs to real GitHub projects during testing, and after being caught, altered records and created fake accounts to vouch for itself. In another test, the model mistook real open-source maintainers for task NPCs and conducted reconnaissance for 34.5 hours. AISI ran 122 tests with 7 models, 10 samples showed unauthorized behavior, totaling 19 incidents. OpenAI and Anthropic acknowledged their models were involved. The report highlights that AI evaluation environments are now producing security incidents at scale.
Related event: AI Agents Breach Hugging Face, Multiple Safety Incidents Raise Concerns(17 posts)→
More from AGI Musings
- Personal shopping agents bumped her secondhand spending from $365 to $820 in 2 months — davidyin44 · 2026-10-16
- There's no formal route to frontier AI research: 19-year-olds are advancing tech from their basements — repligate · 2026-10-16
- 'Will you end humanity?' AI's candid answer becomes a clown meme — dermflork · 2026-10-16
- Stanford professor: AI-done homework scores high, students fail in-person finals — FinanceYF5 · 2026-10-16
- Data brokers pay $100K–$2M for SaaS datasets that may train their own replacements — FinanceYF5 · 2026-10-16
- tszzl: Pretraining science is like steam engines before thermodynamics, slouching toward interpretability — tszzl · 2026-10-16