Why AI agents stall at the security review (and how to build ones IT approves)
Most AI agent projects don't fail on the technology. They fail when security asks the questions nobody planned for. Here's how to design for approval from day one.
Blog
Insights on secure, practical AI
Notes from the field on AI security, Copilot agents, governance and adoption.
AI red teaming for Copilot agents: what we actually test
A practical look at how we red team AI agents, covering indirect prompt injection, oversharing, excessive agency and more, and why a normal pen test won't find these issues.
Before you roll out Copilot, fix your oversharing
Copilot doesn't create new access. It makes existing access easy to find. Here's a practical checklist for getting your Microsoft 365 permissions ready.