Jailbreak-friendly researcher publicly calls model abliteration project harmful
AaronBergman18 · x · 2026-10-22
Aaron Bergman publicly urged @abliterationai to stop its model abliteration work: "I think this is bad and you should stop" — stressing that even as someone inclined to jailbreak models for his own use, he believes making it cheap and easy for the worst actors to strip safety guardrails crosses a line.
More from Safety
- AI assistant dot flags leaked OpenAI and X credentials sitting in the author's Railway project — JasonBotterill · 2026-10-22
- Agent security matures: first CVE-numbered privilege escalation ties agent risk to training conditions — krishnan · 2026-10-22
- PHP2Shell bugs were exploited in the wild days before patches landed — Wired AI · 2026-10-22
- GUDEA's AI traced 25,000 doxxing posts across platforms for NYT front page — nordicinst · 2026-10-22
- Node.js high-severity CVE-2026-111497 enables RCE via Brotli decompression — andersonbcdefg · 2026-10-22
- VX-One: Calibrated decision heads deliver sub-2ms perimeter attack triage against zero-day drift — Juan S. Santillana · 2026-10-22