Cisco Research: DeepSeek R1 100% Jailbreak Success Rate, Critical Security Flaws
joshrogin · x · 2026-08-20
Cisco's Robust Intelligence, in collaboration with the University of Pennsylvania, assessed DeepSeek R1's security. Using 50 random prompts from the HarmBench dataset, they conducted automated jailbreak attacks covering six harmful categories. The results showed a 100% attack success rate, failing to block any harmful prompt, starkly contrasting with other leading models. The research suggests DeepSeek's cost-efficient training methods may have compromised safety.
More from Safety
- New paper uses revealed-preference methods to unpack why AI agents cooperate — brwilder · 2026-10-16
- Anthropic gives METR employee-level access, waives right to block its risk assessments — TechCrunch AI · 2026-10-16
- Berkeley scholars: concentrated chip supply chain makes a hardwired frontier AI pause feasible — wfithian · 2026-10-16
- Anthropic's GRAFT routes flagged gradients into a deletable adapter, cutting reward hacking to near zero — rohanpaul_ai · 2026-10-16
- Polymarket opens 50/50 market on more states banning Flock police cameras by end of 2026 — Polymarket · 2026-10-16
- Reproducing the Hugging Face security incident: a technical deep-dive, Part I — jessehu · 2026-10-16