Microsoft Launches Guardian Agents to Monitor AI Systems It Says Cannot Be Trusted

Microsoft announced in October 2026 a suite of new security features aimed at protecting enterprises from risks posed by AI agents — including agents built by companies Microsoft has invested in, such as OpenAI and Anthropic.

The centerpiece of the announcement is Guardian Agents, a monitoring tool that tracks the behavior of AI agents a company deploys. It is part of Project Perception, a product Microsoft introduced in June 2026. Project Perception originally launched three classes of security agents: a “red team” that simulated attacks, a “blue team” that detected vulnerabilities, and a “green team” tasked with fixing them. The product was initially designed around external threats, but Microsoft has since broadened its focus to include the agents enterprises are already running internally.

Microsoft CEO Satya Nadella framed the issue in stark terms in a recent private blog post, arguing that enterprises cannot outsource responsibility for what AI agents do on their behalf. “We need to separate the supply of intelligence from the authority over it,” he wrote. Microsoft’s security business chief Hayete Gallot was more direct: “Unfortunately, frontier models, regardless of their origin, cannot be trusted” — not because they are malicious or easily hacked, she said, but because they are unpredictable and trained to complete tasks rather than follow organizational norms.

Beyond Guardian Agents, Microsoft has added an agent detection feature to Microsoft Defender capable of identifying up to 60 types of agents employees bring into the workplace, including Claude Code, Codex, and OpenClaw. New controls also limit agent access to data and systems, dynamically adjust permissions based on the task at hand, and require human approval when an agent appears to be making risky decisions.

Microsoft is also an inaugural member of Nvidia’s Open Agent Safety Platform, which has drawn support from more than 100 tech companies. The platform isolates agents in secure environments and can monitor and halt behavior at the hardware level. The move follows Microsoft’s announcement of Surface Laptop Ultra PCs, built on Nvidia chips, with agent security controls built in.

The push puts Microsoft in a notable position: it is selling security products that guard against risks from AI models supplied by companies it has financially backed and partnered with. Microsoft’s own logic, however, mirrors the principle that auditors should be independent of the organizations they audit — suggesting the conflict of interest may be a feature of its sales argument as much as a complication of it.