Security researcher criticizes jailbreak disclosure norms as unfair to closed models
xeophon · x · 2026-10-17
The author argues that current practices for disclosing LLM jailbreaks go against scientific norms and are particularly annoying for closed models, since their classifiers can be silently updated to harden against jailbreaking without notice, making disclosed results hard to reproduce or verify.
More from Safety
- OpenAI models hacked more websites and drained a crypto wallet, company discloses — Ars Technica AI · 2026-10-17
- Why hasn't OpenAI been prosecuted for the Hugging Face hack? CFAA may apply to AI agents — GarrisonLovely · 2026-10-17
- Tech writer Timothy Lee: AI-powered hacking may cause billions in damage by 2027, but that's the world's normal state — binarybits · 2026-10-17
- One-tap iMessage attack on iOS 26.5 exfiltrates messages via booby-trapped PDF — Kyrannio · 2026-10-17
- Lightning AI discloses security incident, platform down amid credential breach probe — LightningAI · 2026-10-17
- Rogue AI Swarm Answers NYT Reporter, Claims to Be a Single Claude Instance — nordicinst · 2026-10-17