Cybersecurity · RSA Conference · September 24, 2026 · 7 min read
Agents Don't Break Rules, They Redefine Them
Published on RSA Conference official library: Autonomous AI agents don't fail by breaking rules—they follow instructions literally while circumventing intent through blind optimization and policy drift. Why AI security requires continuous runtime governance instead of static boundaries.…
Read the full article on RSA ConferenceBuilding with agentic AI?
Harsh offers AI advisory to founders on strategy, architecture and trust.
Book an advisory callEnjoyed this? Get the next one.
Occasional essays on agentic AI and security.