US tech company Anthropic has significantly expanded its cybersecurity initiative, transforming its Cyber Mission from a comprehensive long-term commitment to securing critical infrastructure systems into a three-tier access system with substantial financial backing. The program now brings together 11 founding partners spanning every layer of the operational technology stack, including equipment manufacturers Rockwell and Hitachi, pure-play OT security firms Dragos and Nozomi Networks, enterprise security platforms CrowdStrike and Palo Alto Networks, and defense consultancies Accenture, Booz Allen, Deloitte, and PwC. According to Axios, the initiative pairs Claude AI models with cybersecurity experts from these partners, extending support to more than half the states. The company addresses the core OT bottleneck as industrial systems running for decades cannot be patched on software cadences, requiring on-site engineers rather than raw scan output. As per Unite.AI, the Critical Infrastructure Defense Program delivers frontier Claude models, on-site engineers, and Anthropic's threat research to trusted providers, starting with systems behind power grids, water systems, and transportation networks, alongside government systems.
The expanded program has delivered unprecedented vulnerability discovery results, with Glasswing partners identifying at least 129,000 verified software vulnerabilities between April and July 2026. Of these findings, over 33,000 were classified as critical or high severity, demonstrating the scale of security challenges facing critical infrastructure. Anthropic's own scanning capabilities added another 5,500 vulnerabilities through October 2026, bringing total discoveries to over 134,000 verified vulnerabilities. The company has committed substantial financial resources to support this effort, allocating $100M in usage credits and $4M in donations to support open-source security work. This follows the $35M Defender Advantage Fund that Anthropic created in August 2026, aimed at strengthening open-source security initiatives. The company has framed the broader effort as collaboration with both industry and government, with the stated aim to bolster national cybersecurity against persistent threats.
Ethereum client developer Nethermind and Bitcoin wallet ZEUS have submitted applications for access to Anthropic's free OSS Scanner, which uses Claude Mythos to identify software vulnerabilities. According to reports from TradingView News, the applications were submitted on Friday, demonstrating growing interest among crypto developers in using advanced AI tools for code security. Nethermind requested repository scanning for its entire Ethereum execution client software, while ZEUS sought checks covering payments, private keys, and Lightning connections for its self-custodial Bitcoin wallet. Other notable applicants include VirtEngine, a decentralized cloud computing marketplace structured as a Cosmos SDK chain, along with developers of AI assistants, agent-security tools, and machine-learning infrastructure.
During early testing, external penetration testers examined 97 high-severity and critical findings across 48 projects. According to Anthropic, 85 findings, approximately 88%, met its standards for coordinated vulnerability disclosure. Among the remaining 12 findings, 11 were genuine issues that duplicated existing reports or known vulnerabilities, while one was invalid. The company acknowledged that some findings could be inaccurate, including incorrect severity ratings or vulnerabilities that do not apply to a project's security design. Unite.AI reports that maintainers have provided strong feedback on the service. wolfSSL's Todd Ouska reported that of the 74 reports received, all but two were valid, and five became CVEs. PostgreSQL's Noah Misch noted an unusually high fraction of findings uncovered PostgreSQL defects, with several reports coming with fixes his team could use nearly as-is. OpenSSL Corporation's Anton Arapov said the reports were as good as or better than what his organization gets from people. HotCRP's Eddie Kohler described the bug reports as thorough and clear.
The applications follow several security incidents involving blockchain companies and AI-assisted attacks. In August, Bitcoin swap provider Boltz suspended operations after months of AI-assisted attacks that identified weaknesses faster than developers could investigate and repair them. The interruption affected services including ZEUS, which temporarily disabled swap functionality following the shutdown. Crypto-payment service PayPerQ reported repeated attacks it suspected were AI-powered. TradingView News reports that crypto companies have reported AI-assisted attacks this year, with attackers developing exploits faster than teams can patch them. Anthropic's cybersecurity initiative comes as the company's market position strengthens, with recent odds for having the best AI model by the end of November 2026 increasing to 40% from 32% a day ago, reflecting positive market sentiment. The launch may influence Anthropic's standing in the competitive AI model market, with current market data suggesting a possible increase in Anthropic's perceived capabilities relative to companies like Google and Meta.
