Irregular is testing the cybersecurity capabilities of AI models, including Anthropic’s Claude and OpenAI’s ChatGPT.

Agentic AI functions like an autonomous operator rather than a system that is why it is important to stress test it with ...