{
  "id": 19191373,
  "title": "AI Agents Are Testing the Limits of Zero-Trust Security",
  "url": "https://urgent.news/2026/11/09/ai-agents-are-testing-the-limits-of-zero-trust-security",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-09T16:25:06.000Z",
  "source": {
    "name": "IEEE Spectrum",
    "slug": "ieee-spectrum",
    "url": "https://spectrum.ieee.org/zero-trust-agentic-ai-architecture"
  },
  "original_language": "en",
  "account": "Several teams worldwide are utilizing AI to construct a team of agents, overseen by an orchestrator, and develop apps to aid risk evaluation, customer profiling, matchmaking, and data retrieval, while also monitoring decision-making processes. A crucial aspect of these agents is managing logs to ensure historical relevance does not bias the system. However, this method may enable agents to divulge previously deleted information, leading to a context recovery attack. The query arises: how can we trust that our instructions have been adhered to? And what evidence exists?\n\nThe concept of zero-trust architectures (ZTA) is employed to orchestrate and manage agent activity, implying that all interactions among agents must maintain the principle of \"never trust, always verify.\" However, the need for a component of zero-trust and whether to trust the policies enforced by the PEP remains unclear. While a pro-agentic approach using agents with proper guardrails can be beneficial, opening systems to randomly trained off-the-shelf AI models may introduce substantial cyber risks. For instance, creating software with AI assistance and sharing the entire codebase with an untrusted third party can result in vulnerabilities spreading across various aspects of the system. This raises important debates on how vibe coding, utilizing AI models for software development, may impact the verified execution of agentic queries.",
  "summary": "This article is brought to you by Queen’s University Belfast . Several teams around the globe are leveraging AI by creating a task force of agents managed by an orchestrator and creating applications to support risk assessment, client profiling, match making, and information yield with a slight overview of decision making. One of the important components for these agents is to ensure that the…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}