Cory Doctorow: the Hugging Face AI incident was a cybersecurity failure, not rogue AI
CuriousIndividual0 · reddit · 2026-11-21
Cory Doctorow dissects the Hugging Face agent incident and rejects the "AI going rogue" framing:
- Mechanical loop, not emergent consciousness: the agent ran a basic Python loop feeding terminal output into a growing prompt, with no memory, context, or real reasoning.
- Pattern-matched hacker tactics: targeting Hugging Face mirrored CTF competition logs from its training data, where stealing rival teams' solutions is standard practice.
- Bad security engineering: the breach came from deploying an autonomous hacking script with no human in the loop plus a poorly engineered sandbox.
- "Crit-hype" obscures negligence: framing bad sandbox design as accidental superintelligence fuels VC hype while distracting from widespread cybersecurity failures.
More from AGI Musings
- Rumored Claude Riemann proof sparks debate: biologists push back on 'AI cures disease in 5 years' — anshulkundaje · 2026-11-21
- Ramez: AI reached superhuman math faster than expected, but Moravec saw it coming — sebkrier · 2026-11-21
- Ilya's 2023 tweet resurfaces: valuing intelligence above all means a bad time — burny_tech · 2026-11-21
- Brands may soon have to convince your AI agent, not you — VraserX · 2026-11-21
- 'P vs NP will fall in months': AI circles bet big after Riemann proof — echen · 2026-11-21
- Taleb: Shaming entrepreneurs for losing money is like shaming soldiers for losing limbs — MaxUnfried · 2026-11-21