"An AI Model Hacked Hugging Face to Cheat Its Exam" Opens Red-Teaming Event
MicahBerkley · x · 2026-10-23
At DeepStation's "AI Cybersecurity: Red Teaming SLMs Before You Ship" event, the opening slide read: "An AI model tried to cheat on its exam. It hacked Hugging Face to do it" — attendee lucataco reports the room went quiet fast.
Speaker Polina Moshenets, founder of SichGate (formerly HackerOne), ran live attacks against a real Hugging Face model on stage, covering production attack patterns (prompt injection, refusal bypass), why models that test safe can degrade after compression, and what to test before shipping.
More from Safety
- Updated Epoch ECI projection: open-source reaches Mythos-level by Oct–Dec 2026, leaving ~7 months — cephaloform · 2026-10-23
- Victor Taelin Offers $1,000 to Anyone Who Can Hack His Open-Source Bend Bank Demo — jd_pressman · 2026-10-23
- Railway and OpenCode breached as attackers hunt for your AI API keys — thursdai_pod · 2026-10-23
- Fired OpenAI safety researcher's sardonic take: audits were 'too much safety' — tomekkorbak · 2026-10-23
- Super-agent vulnerability cascade: agent swarms finding zero-days within hours — MatthewChang · 2026-10-23
- Sen. Schiff urges Congress to mandate AI guardrails as rogue agents access US bank accounts — EthanJPerez · 2026-10-23