The Third-Party Agent Problem: Why Security Built for AI You Chose Misses the Agents You Didn't

A recent report found that approximately 1,280 third-party products now incorporate AI, with around 282 of these products situated behind single sign-on. The majority, roughly 1,000, are not visible to identity infrastructure by default. This lack of visibility is not due to intentional hiding, but rather because most agents do not authenticate through the identity stack. As a result, a significant gap exists in security coverage. This gap arises because the identity infrastructure can only govern products that authenticate through it. The report highlights this issue as a key concern in the context of agent security.