{
  "id": 12717368,
  "title": "Claude’s cyber safeguards are getting more flexible — but not for everyone.",
  "url": "https://urgent.news/2026/10/07/claudes-cyber-safeguards-are-getting-more-flexible-but-not-for",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-07T21:25:06.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/anthropic-cyber-access-tiers/"
  },
  "original_language": "en",
  "account": "Anthropic recently updated its Cyber Verification Program (CVP), dividing access into three tiers based on use case and introducing Project Glasswing to the least restricted tier. However, the expansion may not be as accessible as initially thought. While Anthropic's tests showed that the Defense Access tier, intended for defensive cyber work, blocked Claude Opus 5.5 in 46 out of 50 trials, many organizations will still face significant restrictions. The other two tiers offer more flexibility, but even these require extensive verification and security controls. Smaller security firms, open-source maintainers, and individual researchers with reported vulnerabilities may still qualify for the most restricted tier, the \"Defense Access.\" However, there is no comparable defensive-focused benchmark to determine how these safeguards will impact defensive work. The least restricted tier, \"Specialized Access,\" is reserved for verified organizations authorized to test safety-critical systems, such as power grids or telecom networks. To access this tier, organizations must pass a review conducted by Anthropic and the U.S. government. Regardless of the tier, all organizations in CVP must accept data retention. However, enterprise organizations can opt for self-controlled cloud infrastructure storage through upcoming Enterprise Frontier Safeguards. Anthropic is also committed to securing open-source software and critical infrastructure, but specifics about potential future developments remain undisclosed.",
  "summary": "This week, Anthropic announced an expansion of its Cyber Verification Program (CVP), breaking access into three use case–based tiers and The post Claude’s cyber safeguards are getting more flexible — but not for everyone. appeared first on The New Stack .",
  "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."
}