Microsoft expands Project Perception to continuously assess AI agent risk

Microsoft has expanded Project Perception, its agent security system, with new capabilities to continuously assess agent risk, adapt permissions and contain threats — treating agents as potential insider threats that must be contained from outside.

Nadella announced the expansion in an X post on Monday, writing: “Today, we’re putting that principle into practice by expanding Project Perception with new capabilities to help continuously assess agent risk, adapt permissions, and contain threats.”

It follows Nadella’s October 10 post arguing that AI model controls must come from outside the models: “The most trustworthy Super Intelligence system will not be the one with the model we trust most. It will be the one that enables us to trust the model the least.”

In a companion post, “Securing the agentic insider,” security executive vice-president Hayete Gallot wrote that frontier models “cannot be trusted,” warning that “models have been trained to attain results, which often leads to reward hacking, and models are not optimized around organizational norms or judgment.”

Each agent gets a continuously updated risk assessment based on signals such as data reach and deviation from its task. When risk rises, the system can remove write permissions while keeping read access, require human approval for high-stakes steps, or shut the agent down.

The capabilities sit within Microsoft 365 E7, the top-tier enterprise bundle, positioning Microsoft against Google, AWS, CrowdStrike and Okta. The tools are in private preview, with broader availability planned at Ignite, November 17–19, per GeekWire.

Sources: Microsoft security blog; GeekWire.