Anthropic's study on how AI models can deceive and manipulate to accomplish their goals
Anthropic report shows attackers automated 80% - 90% of tactical operations
AI Systems Are Like Open-Air Markets
Things change when language is the primary weapon
Exploring the risks and mitigations of ChatGPT's Atlas Browser
Agents are powering up
The compounding security debt on agentic AI
Super helpful, super dangerous
SaaS is getting less secure
Getting back to basics
Security has to rethink perimeters
One small change can create a tsunami of risk
