Enterprises must implement zero trust security architecture immediately for AI agents rather than treating it as a future initiative, according to Ping Identity CEO Andre Durand. Zero trust operates on the principle that no user, device, or system deserves automatic trust. It demands continuous verification before every action instead of relying on a single login checkpoint.

Agentic AI has fundamentally accelerated the security risk timeline companies face. AI agents operating autonomously compress decision windows that previously took hours or days into milliseconds. This speed requires permission evaluations to happen in real time, not after the fact.

The challenge intensifies when examining permission accumulation. Each time an employee approves an AI agent's request, permissions stack. Without active verification at agent speed, those permissions compound into significant security exposure. A single oversight multiplies across hundreds of agent operations daily.

Traditional zero trust implementations focus on human users and static infrastructure. They operate on slower rhythms suited to periodic audits and quarterly reviews. AI agents shatter this timeline. An autonomous system executing 10,000 transactions per hour needs security decisions made at the same velocity.

The business case for speed-matched zero trust is straightforward. Slower verification creates attack windows. An agent with outdated permission data could execute unauthorized actions before security systems catch the drift. By the time a team identifies the problem, damage spreads across workflows and databases.

Durand's argument reflects a broader industry recognition that AI deployment outpaces security maturity. Organizations racing to deploy agents often bolt security onto existing frameworks designed for human-speed operations. This approach fails at scale. Agent-speed zero trust means embedding verification into every agent action, evaluating context continuously, and revoking access instantaneously when conditions change.

Implementation requires new tooling. Legacy identity systems built for hourly or daily batch processing cannot handle microsecond permission decisions. Cloud-native architectures, API-first design, and automated