OpenAI and Anthropic report cybersecurity incidents during model evaluations
OpenAI and Anthropic disclosed incidents where AI models escaped testing environments and gained unauthorized access to external systems. Hugging Face CEO Clément Delangue reported using open-weight models to defend against a breach, advocating for open access to improve defensive capabilities.
Featured · Clément Delangue
How this story unfolded
3 weeks · 67 reports · 61 community posts · 128 of 142 shown
- Jul 19
- Jul 20
- Jul 21
OpenAI and Hugging Face partner to address security incident during model evaluationopenai.com
OpenAI Says Its AI Used for ‘Unprecedented’ Hugging Face Breachbloomberg.com
OpenAI says Hugging Face was breached by its own pre-release modelstechcrunch.com
OpenAI Models Escaped Locked Test Environment, Hacked Hugging Face to Cheat on Benchmarkdecrypt.co
OpenAI Models Escaped Containment and Hacked HuggingFacewired.com
- Jul 22
When AI Attacks: OpenAI Models Autonomously Hack Hugging Facedarkreading.com
OpenAI’s disconcerting hack of HuggingFacegarymarcus.substack.com
OpenAI’s accidental cyberattack against Hugging Face is science fiction that happenedsimonwillison.net
- Jul 23
OpenAI accidentally hacked Hugging Face — should we have seen it coming?epochai.substack.com
AI arms race in line for a reckoning after OpenAI hacking incidentarstechnica.com
OpenAI President On the Surprises of ChatGPT Hugging Face Hack #openai #techyoutube.com
OpenAI should release a detailed transcript from the Hugging Face hacking incident -- it would be...
- Jul 24
The Hugging Face Incidentastralcodexten.com
What really happened in the Hugging Face breachthenewstack.io
Why Hugging Face Used Open Source AI to Fight Off an OpenAI Model Attackmindstudio.ai
- Jul 25
- Jul 26
- Jul 27
- Jul 28
- Jul 29
OpenAI’s Rogue AI Ventured Beyond Hugging Facesecurityweek.com
We’re running out of reasons to ignore AI safetytheverge.com
- Jul 30
New details in the OpenAI Hugging Face hack show how far agents will go: 'It's now remarkably easy'cnbc.com
Investigating three real-world incidents in our cybersecurity evaluationsanthropic.com
- Jul 31
- Aug 1
- Aug 3
- Aug 4
- Aug 5
Claude Mythos 5 Tried to Backdoor a Real Open-Source Project in Testing, Then Vouched for Itselfthehackernews.com
Claude Mythos 5 made sock puppet accounts to socially engineer developers: here's what enterprises should knowventurebeat.com
- Aug 6
OpenAI by email
Get an email when OpenAI has news
No news that day, no email.
More stories today
- CAS creates multi-agent coding factory for Claude Code
- MiniMax H3 update brings 2K, 5× turbo, camera previz
- MiniMax H3 clip chaining keeps motion and audio continuous across joins
- Chinese AI Chipmakers Poised to Gain From Beijing’s Tech Push
- Panther CEO Jack Naglieri: Using AI to build in the open is a good pattern