Physix Frontier · News Briefing Card (IT Home · Oct 7, 2026)

Anthropic expands security program, over 130K flaws found

KEY FACTS

  • Anthropic unveiled its new Cyber Verification Program (CVP) on October 7, consolidating two earlier security initiatives.
  • Partners in the Glasswing program verified at least 129,000 software vulnerabilities between April and July.
  • Anthropic's own open-source scanning uncovered an additional 5,500 vulnerabilities from April to October.
  • More than 33,000 of those vulnerabilities were rated critical or high severity.
  • The new program has three tiers - defense, red team, and specialized - all of which can use models such as Claude Opus 5.5.

KEY DATA

129,000Partner-verified vulnerabilities
5,500Additional vulnerabilities found by open-source scanning
33,000Critical or high-severity vulnerabilities
3Number of program tiers

PHYSIX OBSERVATION

AI is shifting from a code-writing tool to attack-and-defense infrastructure for finding vulnerabilities. By opening its most powerful models in tiered access based on qualifications, Anthropic is essentially building a gate between security capability and abuse risk. For the industry, vulnerability discovery efficiency has been raised sharply, but whether defenders can keep pace with remediation is the real test.

Source: IT Home report