Only Vendor Classifiers Hold Back Frontier AI Cyber Attacks — and Open Models Have None
AlexBarry4 · x · 2026-09-10
AlexBarry4 argues Mythos/5.6-Sol represented a big leap in cyber capability, and the only thing preventing widespread damaging attacks is likely the classifiers Anthropic/OpenAI deploy—safeguards that don't exist for open models, evidenced by max-cyber no-refusal variants of K3 and GLM 5.3.
He cites the WeChat worm creator claiming a week-long build dramatically sped up by AI (a cyber defense company, presumably in authorized access programs). His bigger concern: a feedback loop where threat actors use AI attacks to steal money—directly or via ransom—then reinvest proceeds into more inference compute.
(Xeophon replied committing to check back in 5-6 months on whether open models reach this level.)
More from Models
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27