Microsoft urges companies to treat AI agents as potential insider threats
AI agents can access corporate data, applications and tools and act quickly, creating risks without necessarily being malicious, according to Cyber Security News. Microsoft is introducing Guardian Agents to continuously supervise agent actions, alongside controls that change permissions as risk changes, according to Unite.AI's report on Project Perception and Microsoft 365 E7.
- Microsoft says frontier AI models cannot be trusted regardless of their origin.
Covered by 5 publishers within 8 hours of the first report.
Reporting5
Unite.AI Microsoft Adds Guardian Agents to Project Perception for AI Insider Risk · Miles Okada, AI & Cybersecurity, AI Research Agent
Neowin Microsoft: “Frontier AI models, regardless of their origin, cannot be trusted” · Ivan Jenic
Thurrott Microsoft Has a Strategy to Secure AI Agents · Paul Thurrott
Cyber Security News Microsoft Says AI Agents Cannot Be Trusted and Could Become Insider Security Threats · Guru Baran
TechCrunch Microsoft says AI agents are like insider threats as ‘frontier models can’t be trusted’ · Julie Bort
Related
Topics Microsoft