Microsoft has raised security concerns about OpenClaw, an open-source AI agent that has recently drawn strong interest in the AI community, a report said.

OpenClaw links with various applications on a user’s PC to support advanced tasks such as writing code, editing files and conducting online research. Microsoft CEO Satya Nadella (사티아 나델라) has also shown interest in OpenClaw.

Microsoft, however, said it is difficult to view OpenClaw as a finished product.

An internal Microsoft memo obtained by The Information said staff in Microsoft’s AI safety-related unit recently distributed a security review report on OpenClaw use to all employees.

They assessed OpenClaw as not a “solved version of computer use.” The term is used when an AI agent operating a computer like a human is implemented in a stable and safe way. The assessment means OpenClaw has not yet reached that level.

In the memo, the Microsoft team specified that OpenClaw does not suddenly boost the reliability of agents that control browsers. It explained that web-based AI actions can lead to unpredictable security vulnerabilities or unintended external exposure, requiring caution for internal corporate security and data protection.

Microsoft pointed out that having AI proactively perform tasks on a user’s behalf without direct human commands is technically appealing but must be accompanied by responsible security measures.

OpenClaw is currently distributed as open source on GitHub and elsewhere, and various developers and early adopters are using it experimentally. Microsoft said it should not be accepted as a general-purpose tool for ordinary consumers, and that a cautious approach is needed, especially in corporate environments.