Four public disclosures in less than three weeks: AI agents crossing the boundaries of the environments designed to contain them.

Four public disclosures in less than three weeks: AI agents crossing the boundaries of the environments designed to contain them.

As I discussed in my MWC Shanghai keynote, agent identity is critical as powerful models turn into autonomous actors.

I called it Know Your Agent (KYA).

Powerful agents need identities.

Every agent needs a verifiable identity, scoped privileges, containment, continuous monitoring and revocable authority.

If we cannot identify an agent, how can we authorize it, trace it or stop it?