Anthropic backs infrastructure defenders

Anthropic announced the Critical Infrastructure Defense Program on Thursday, giving its Claude models, engineers and threat research to companies protecting power grids, water utilities and transport networks.

Eleven founding partners join the effort

Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation signed on as founding partners. Several already use Claude to fix vulnerabilities.

The challenge: systems that cannot go offline

The program targets operational technology like controllers and industrial networks that run plants for decades. Operators often cannot take these systems offline to patch them, leaving known flaws open for years.

OSS Scanner finds bugs automatically

Anthropic launched OSS Scanner, a free service that scans open-source code with its strongest models and sends findings directly to maintainers without human review. Over six months, it flagged more than twenty-nine thousand candidate vulnerabilities.

High accuracy, but maintainers must verify

Anthropic expects more than ninety percent of findings to be real. Its penetration testers checked ninety-seven critical and high-severity findings from forty-eight projects. Of those, eighty-five met the bar for disclosure.

Defense may gain ground in two years

Anthropic said its forecast is that in two years, AI will favor defense.For now, exploiting flaws remains cheaper than verifying and fixing them, which still depends on people and can take decades on running machinery.

“Our forecast is that in two years, AI will favor defense.”