A little while ago I demonstrated something fairly simple. An attacker-controlled HTTP User-Agent value was ingested into Microsoft Sentinel as part of normal security telemetry. Security Copilot then consumed that telemetry while analysing the incident. The User-Agent contained an indirect prompt …

This is a personal blog. All opinions are my own - not my employer’s. There’s apparently a new category of security incident called AI escape. Except there isn’t. That phrase might be useful shorthand for headlines, and some of the incidents behind it are genuinely serious, but I think …

Hero image generated by ChatGPT This is a personal blog. All opinions are my own and not those of my employer. From Clawdbot to GAINet: When Agent Experiments Outrun Accountability This post didn’t start as a philosophical musing about AI. It started with a very practical, very familiar security …