Although this month we planned for a talk from Phill Fox, at short notice he was taken ill, so we switched to the talk planned for next month about Rogue AI (unfortunately I didn’t get chance to rehearse .. hopefully it wasn’t too bad!). The good news is that Phill has agreed to do the talk on the 17th Nov from CERN, so I’ll do a repeat invite to the groups next month.
We covered a brief re-cap of Large Language Models (LLM) and Agentic AI, followed by some events that led up to the main incident. We looked at the DseWiki and RubyGems attacks before looking at the OpenAI/Hugging Face incident and the Australian medicare attack. Finally we look at some conclusions of the events and discussed possible futures.
The slides (https://docs.google.com/presentation/d/1s8cNt3WpvDykqkgGx1i8gxcEpakM-Sihnesww19BzHA/edit?usp=sharing) were :
The recording (https://youtu.be/a11trRhvNis) was :
The references were :
- https://en.wikipedia.org/wiki/OpenAI%E2%80%93HuggingFace_incident
- https://openai.com/index/hugging-face-incident-and-the-road-ahead/
- https://www.irregular.com
- https://en.wikipedia.org/wiki/OpenAI_rogue_agent_breach_of_Medicare
- https://www.darkreading.com/cyberattacks-data-breaches/openai-agents-wiki-site-hugging-face-attack
- https://collusion.wiki/?ref=explainx
- https://rubyhack.ai
- https://www.bbc.co.uk/news/articles/ckgwy1k42w4o
- https://www.forbes.com/sites/siladityaray/2026/09/28/anthropic-ipo-prospectus-warns-its-ai-could-pose-existential-risks-to-humanity/
- https://youtu.be/87DyyMV0kCY
- https://metr.org/hugging-face-incident-report-aug-2026.pdf
- https://en.wikipedia.org/wiki/AI_alignment
- https://metr.org/blog/2026-06-26-gpt-5-6-sol/
- https://github.com/webpro255/awesome-ai-agent-attacks
- https://www.netflix.com/title/82792555
- https://www.netflix.com/gb/title/82772145
Note that the Hugging Face CEO isn’t that happy about the latest netflix documentary – see https://www.businessinsider.com/hugging-face-ceo-criticizes-netflix-documentary-ai-gone-wild-2026-10 for his comments.