Microsoft expands Project Perception to contain anomalous AI agents
Original title: AI Agents as Insider Threats: Inside Microsoft's Bet on a Unified Security Stack
New capabilities discover agents, assign identity and permissions, score risk, and can cut off access.
Microsoft added capabilities to Project Perception so teams can discover AI agents in an organization, assign each an identity and permissions, and continuously score risk. When it finds what Microsoft calls an agentic insider, the system can cut off that agent's access. The first public preview, offered through Microsoft Defender in August, coordinated red, blue and green agents and used the MDASH scanner with MAI-Cyber-1-Flash. Experts cited by Dark Reading warn that using agents to police other agents may increase risk.
Unconfirmed: Whether agent-on-agent monitoring reduces risk is disputed by experts quoted in the article.
- Guardian Agents are intended to monitor and contain anomalous agents.
- The August preview focused on vulnerability management through MDASH, still in preview.
- MDASH uses several models, including MAI-Cyber-1-Flash.
- Microsoft says agents should be treated like privileged insiders.
- Vendors
- Microsoft
- AI models
- MAI-Cyber-1-Flash
From the original article
11 min read at darkreading.com
Microsoft's Project Perception adds Guardian Agents to monitor and contain anomalous AI agents, but experts warn that policing unpredictable agents with other agents may multiply — not contain — the risk.
By Jeffrey Schwartz · darkreading.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.