[{"data":1,"prerenderedAt":30},["ShallowReactive",2],{"nr-en-anthropic-claude-cyber-verification-program-expansion":3},{"slug":4,"title":5,"dek":6,"date":7,"time":8,"publishedAt":9,"updated":10,"updatedAt":10,"dateFmt":11,"updatedFmt":10,"kind":12,"tier":13,"author":14,"authorName":15,"topics":16,"tracker":22,"trackerLabel":23,"headlineStat":24,"image":25,"ogImage":26,"imageAlt":5,"csv":10,"minutes":27,"words":28,"html":29},"anthropic-claude-cyber-verification-program-expansion","Anthropic Opens Claude to Security Researchers – With Reduced Safety Filters","The AI provider expands its Cyber Verification Program with a structured three-tier access model. Security teams gain access to Claude's most capable models with loosened safety guardrails for penetration testing and vulnerability research.","2026-10-07","14:44","2026-10-07T14:44:00+02:00","","October 7, 2026","news","standard","ideal-syka","Ideal Syka",[17,18,19,20,21],"AI Security","Anthropic","Cybersecurity","Claude","Vulnerability Research","\u002Feu-ai-act-fahrplan","AI Act & Governance","129,000 vulnerabilities in 4 months","\u002Fnewsroom\u002Fimg\u002Fanthropic-claude-cyber-verification-program-expansion.webp","\u002Fog-nr\u002Fanthropic-claude-cyber-verification-program-expansion.en.png",2,495,"\u003Cp>Anthropic is making good on its strategy to institutionalize AI security research. The company is expanding its \u003Cstrong>Cyber Verification Program (CVP)\u003C\u002Fstrong> to significantly more security professionals, granting them access to Claude's most capable models with reduced safety filters—specifically for tasks like vulnerability discovery, malware analysis, incident response, and penetration testing.\u003C\u002Fp>\n\u003Ch2>The essentials\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Cstrong>Three access tiers\u003C\u002Fstrong>: Defense Access (enterprises, government agencies, universities, critical infrastructure operators), Red Team Access (authorized attack simulations for organizations only), and Specialized Access (aviation systems, power grids, banking infrastructure—reviewed jointly with the U.S. government)\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Early results\u003C\u002Fstrong>: Partners in the predecessor program Project Glasswing found at least \u003Cstrong>129,000 confirmed vulnerabilities\u003C\u002Fstrong> between April and July 2026, with over \u003Cstrong>33,000 classified as critical\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Acceleration effect\u003C\u002Fstrong>: Multiple partners report Claude has accelerated their work by months to years\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Regulated access\u003C\u002Fstrong>: Anthropic reviews applications itself or jointly with the U.S. government—no uncontrolled release\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2>Why safety filters stay on by default\u003C\u002Fh2>\n\u003Cp>Publicly available Claude models block most of this security work—not out of malice, but logic: the same capabilities that help security researchers find vulnerabilities also benefit attackers. Simple tasks like code review or patching known flaws remain possible without special access. But deep penetration testing or malware analysis? Claude normally declines.\u003C\u002Fp>\n\u003Cp>The CVP resolves this conflict through trust verification and transparency: applicants are vetted, access is granted under conditions Anthropic can track.\u003C\u002Fp>\n\u003Ch2>Three tiers for different risk profiles\u003C\u002Fh2>\n\u003Cdiv class=\"tbl-scroll\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Access Tier\u003C\u002Fth>\n\u003Cth>Target Audience\u003C\u002Fth>\n\u003Cth>Use Case\u003C\u002Fth>\n\u003Cth>Key Feature\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>\u003Cstrong>Defense Access\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Security teams, universities, open-source developers, individual researchers\u003C\u002Ftd>\n\u003Ctd>Vulnerability discovery, incident response\u003C\u002Ftd>\n\u003Ctd>Open to individuals\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Red Team Access\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Organizations\u003C\u002Ftd>\n\u003Ctd>Authorized attack simulations\u003C\u002Ftd>\n\u003Ctd>Organizations only, not individuals\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Specialized Access\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Critical infrastructure (aviation, power grids, banking systems)\u003C\u002Ftd>\n\u003Ctd>Testing on high-sensitivity systems\u003C\u002Ftd>\n\u003Ctd>Joint review with U.S. government\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2>The numbers validate the approach\u003C\u002Fh2>\n\u003Cp>The predecessor program Project Glasswing already demonstrates what's possible. Between April and July 2026, program partners found at least \u003Cstrong>129,000 confirmed vulnerabilities\u003C\u002Fstrong> according to Anthropic. Over \u003Cstrong>33,000\u003C\u002Fstrong> were classified as critical or high-severity. Important caveat: these figures come from surveys of a subset of partners—Anthropic estimates the actual impact is at least five times higher.\u003C\u002Fp>\n\u003Cp>Multiple partners reported that Claude accelerated their work by months to years. That's not marketing speak; it's evidence that AI-powered security research actually scales.\u003C\u002Fp>\n\u003Ch2>What this means for enterprises and governments\u003C\u002Fh2>\n\u003Cp>The CVP expansion signals that Anthropic positions Claude as a tool for defenders, not attackers—and is building institutional structures to prove it. For enterprises and government agencies evaluating Claude, this means: if your security team needs access, there's now a clear pathway. The open question remains: how fast and frictionless is approval? And will joint U.S. government review of critical infrastructure access create political friction?\u003C\u002Fp>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fthe-decoder.de\u002Fanthropic-oeffnet-cyber-verification-program-sicherheitsforscher-bekommen-claude-mit-reduzierten-schutzfiltern\u002F\">The Decoder (DE)\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.securityweek.com\u002Fanthropic-introduces-3-tier-cyber-verification-program-for-ai-access\u002Famp\u002F\">SecurityWeek\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fthe-decoder.com\u002Fanthropic-gives-more-security-teams-access-to-claude-with-fewer-safety-restrictions\u002F\">The Decoder\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.heise.de\u002Fnews\u002FAnthropic-buendelt-Cyberprogramme-und-fuehrt-drei-Zugriffsstufen-ein-11478953.html\">heise online\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cem>Editorially owned by \u003Ca href=\"\u002Fen\u002Fautor\u002Fideal-syka\">Ideal Syka\u003C\u002Fa>. Sources and method: \u003Ca href=\"\u002Fen\u002Fredaktion\">Newsroom & method\u003C\u002Fa>. Tips and corrections: \u003Ca href=\"mailto:ai@i6eal.de\">ai@i6eal.de\u003C\u002Fa>.\u003C\u002Fem>\u003C\u002Fp>\n",1791377688053]