Microsoft warns against AI agents and proposes security solutions
In brief
- Microsoft views AI agents as inherently untrustworthy actors, even without malicious intent.
- The company proposes tools like "Guardian Agents" and sandbox environments (MXC) to mitigate risks.
- The announced solutions require expensive subscriptions (e.g., Microsoft 365 E7) and lack deployment details.
Microsoft published a post on October 19 to highlight security risks linked to AI agents. According to the company, cutting-edge models are inherently unreliable because they are trained to achieve goals that may divert them from their intended function. Microsoft suggests measures such as expanding its Perception project, including "Guardian Agents" to monitor agent actions in real time.
Hayete Gallot, Microsoft’s vice president of security, states that agents should be treated as untrusted "insiders." The company details tools like Dynamic Permissions, sandbox environments (MXC), and dynamic risk assessment systems (Dynamic Agent Risk) to limit agent permissions and react to deviations. Nadim Abdo, vice president of Identity and Network Access, emphasizes the need to "trust the model as little as possible" to ensure security.
These announcements raise questions: Microsoft, both judge and jury, sells security solutions while describing risks. The proposed architecture blends deterministic controls with model-based monitoring, adding a new layer of complexity. Companies will need further details on costs and required subscriptions to deploy these tools.
What it changes for you
- CybersecurityCompanies will need to integrate continuous monitoring tools for AI agents, increasing security complexity and costs.
This summary gives you the essentials. The full article, its figures and quotes are on Next.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error