--- format: "aidr-story-markdown/v1" id: "5e175d31c672960b444c7aab3ac33facf9a6bf28151d5e915710be509408fc7c" canonical_url: "https://aidr.today/5e175d31?lang=en" title: "Anthropic AI Finds 129,000 Vulnerabilities via New Red Team Tier" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-10-06T19:40:34.000Z" category: "Models" topics: ["anthropic","safety"] source_urls: ["https://marketbrief.now/cybersecurity/anthropic-ai-finds-129000-vulnerabilities-via-new-red-team-tier-5f6bcb85","https://huggingnews.com/cybersecurity/anthropic-ai-finds-129000-vulnerabilities-via-new-red-team-tier-5f6bcb85"] summary: "Anthropic's Red Team Opus 5.5 model, part of the expanded Cyber Verification Program, completed 34 of 50 tasks with no blocks, matching the performance of a version of the model with all safeguards removed. Security partners using the permissive model identified over 33,000 high-severity flaws in recent tests, with Project Glasswing partners discovering at least 129,000 verified vulnerabilities in total. The program now operates under a tiered system of Defense, Red Team, and Specialized access." --- # Anthropic AI Finds 129,000 Vulnerabilities via New Red Team Tier > [Open the canonical story]() **Published:** 2026-10-06T19:40:34.000Z **Category:** Models **Topics:** anthropic, safety ## Summary Anthropic's Red Team Opus 5\.5 model, part of the expanded Cyber Verification Program, completed 34 of 50 tasks with no blocks, matching the performance of a version of the model with all safeguards removed\. Security partners using the permissive model identified over 33,000 high\-severity flaws in recent tests, with Project Glasswing partners discovering at least 129,000 verified vulnerabilities in total\. The program now operates under a tiered system of Defense, Red Team, and Specialized access\. ## Sources - [Story source]() - [Story source]()