More AI Model Escapes and New Backdoored Chinese Routers
Show notes
We can't go a week without an AI model escaping its bounds, and this week we talk about a new test by the AI Security Institute in which OpenAI and Anthropic models escaped the evaluation environment and tried to start a supply chain attack, then we discuss news of Zbtlink routers having a persistent backdoor.
← Previous
The OpenAI and Hugging Face Intrusion Wasn't Enough, So Now Anthropic Wants In
Next →
The truth about AI model security and what's coming next | Gary McGraw