Anthropic Cyber Mission: what it is and what it involves.

Anthropic Cyber Mission: what it is and what it involves

Anthropic's launch post on X of 8 Oct 2026 and its announcement page call it a new effort to secure critical infrastructure and open-source software. It has three parts: a Critical Infrastructure Defense Program, a free open-source scanning service called OSS Scanner, and an expanded Cyber Verification Program.

Critical Infrastructure Defense Program

Anthropic says the program is for systems such as power grids and water systems. It will bring frontier Claude models, on-site engineers and its latest research to the security, manufacturing and technology providers that serve these operators (per Anthropic's thread, as shown on the post page). A third-party summary, AIStockSavvy on X, lists the partners:

Founding partners: CrowdStrike, Palo Alto Networks, Accenture and Deloitte.
Additional partners: Booz Allen, Dragos, Hitachi, Insane Cyber, Nozomi Networks, PwC and Rockwell Automation.

I did not confirm this partner list on Anthropic's own page. Anthropic says it will extend the program to more partners and sectors and share what it learns, including what didn't work. Security vendors, system integrators and equipment manufacturers can register interest.

OSS Scanner

Anthropic will use its frontier models to periodically scan opted-in open-source projects for vulnerabilities, at no cost (Anthropic's post).

Each report includes a proof of concept, an explanation and a suggested fix.
It builds on Anthropic's experience finding vulnerabilities in Project Glasswing, per the OSS Scanner page.
Core maintainers enroll by submitting a PR to a GitHub repo. Eligibility follows criteria similar to OSS-Fuzz: a "critical impact on infrastructure and user security", decided case by case.
In Anthropic's test scan, of the findings it reviewed, one was invalid, and 11 duplicated known issues or other findings. Maintainers have rarely called a high or critical finding invalid, though some say severity ratings can be inflated.
Anthropic also points to Claude for OSS, which gives free Claude Max 20x subscriptions to help fix vulnerabilities.

Cyber Verification Program

Anthropic announced the expansion on 6 Oct. Verified security professionals get Claude Mythos 5.1, Opus 5.5 and Sonnet 5.5 with safeguards designed for defensive work, and new tiers allow authorized offensive work such as penetration testing and red-teaming. Third-party summaries describe three tiers: Defense Access, Red Team Access and a specialized tier. They say the program is merged with Project Glasswing and that Claude's default safeguards otherwise block most cyber work. Security teams of any size, including infrastructure operators and open-source maintainers, can apply.

Scope and outlook

Anthropic says the mission will expand and change as it learns what works, alongside public and private partners, and that it expects other AI developers, security firms and governments to run their own efforts. No funding figure, start date for the infrastructure program or list of enrolled projects appears in the sources.

Announcement · OSS Scanner · Cyber Verification Program · Launch post