{
  "id": 18893546,
  "title": "Allow, Ask, or Block: The Week Agent Security Moved to the Action",
  "url": "https://urgent.news/2026/11/08/allow-ask-or-block-the-week-agent-security-moved-to-the-action",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-08T08:11:50.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/michaelpetler/allow-ask-or-block-the-week-agent-security-moved-to-the-action-4hc1"
  },
  "original_language": "en",
  "account": null,
  "summary": "The security industry has recently moved towards a more proactive approach in safeguarding AI agents, focusing on monitoring and controlling the actions agents take rather than just the information they access. This shift is evident in the recent releases of three major security vendors: Sysdig AI Defense, CrowdStrike, and Anthropic. All three solutions aim to enforce policies and make real-time decisions on whether to allow, ask, or block actions taken by agents. This change is crucial as it addresses vulnerabilities in agent platforms themselves, such as the critical flaws found in IBM's Langflow OSS. By implementing this new security paradigm, organizations can better protect their systems and data from potential threats posed by AI agents.",
  "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."
}