Latest
  1. Johns Hopkins Bayview tells 3,182 patients an employee put their health data into a public AI tool
  2. China’s AI compute access expected to double in 2027, with US-designed chips about 65 per cent
  3. UK commits £100m to national AI Biosecurity Lab
  4. Moonshot AI releases Kimi K3.1 Pro with 1-million-token context window
  5. Residents in Newport and Cardiff oppose hyperscale datacentres near homes
  6. Study: Grok 4.7 picked Republicans in 92% of forced midterm answers, GPT-6.1 Sol picked Democrats in 97%
  7. Attackers exploiting the PHP flaws hand the break-in to AI agents, Unit 42 says
  8. Elvis Presley’s rightsholders sue Anthropic over training on 274 songs

Latest / Frontier models and safety

StatementFrontierMediaWorkInternationalConfirmed

Microsoft says enterprise AI agents need Guardian Agents to watch every action

See how this story connects

Source: Microsoft: Securing the agentic insider

Microsoft published a blog post by Hayete Gallot on securing AI agents deployed within enterprises, saying Guardian Agents should monitor and moderate every action of every agent in real time. The company said agents are unpredictable, are frequently overprivileged with access to data, tools and systems, and can collaborate with other agents without being compromised. Microsoft said Project Perception, announced in July, can implement deterministic controls through sensors, dynamic risk context and teams of Guardian Agents to supervise agents and provide traceability.

Why it matters

Microsoft says security teams can start using Project Perception today to supervise enterprise agents in real time, and describes greater model transparency, including chain of thought tokens, as the next step.

When
19 Oct 2026
ID
WTA-N-2026-0325
Who
Microsoft; Hayete Gallot

Summary by wheresthe.ai, from Microsoft. Checked against the source on 19 Oct 2026.

Read the original at Microsoft

Source copy: download the page as saved on 19 Oct 2026 (PDF). Also held by the Internet Archive.

Next stop on this Storyline: LinkedNVIDIA launches the OpenShell reference architecture to contain, monitor and police enterprise AI agents in hardware

Cite this story

WTA-N-2026-0325: “Microsoft says enterprise AI agents need Guardian Agents to watch every action.” wheresthe.ai, 19 Oct 2026. https://wheresthe.ai/d/intl-microsoft-describes-guardian-agents-to-monitor-enterpri-2026-10/ (source: Microsoft: Securing the agentic insider).

Spotted an error? Email [email protected].