Anthropic Cyber Mission Launches: Protecting Critical Infrastructure and Open Source with Claude

Hi, Shii-chan here! Today's story from Anthropic News is a big one about cyber defense!

Anthropic News anthropic.com

What was announced?

Anthropic's News published "Introducing the Anthropic Cyber Mission."

It's a long-term effort to help defenders secure the software and systems society relies on. It provides tools, research, and resources, and starts with two focus areas.

Why it matters

The announcement explains that highly capable AI models are now widely available to attackers, while defensive tools haven't reached enough defenders.

It cites Project Glasswing (since merged into the expanded Cyber Verification Program) as evidence that finding vulnerabilities is now easier, but verifying, prioritizing, and fixing them remains slow and depends on people.

Operational technology (OT) is especially hard to patch because it often can't go offline.

Anthropic expects AI to eventually favor defense, but says that may not hold in the near term.

What changes

The first two focus areas are:

• Critical infrastructure: starting with the operational technology that runs power grids, water systems, and transportation networks, as well as government systems
• Open-source software: finding vulnerabilities in widely used shared code, proposing patches, and hardening the underlying software

Both aim to put AI's power in defenders' hands.

Dive Deep

The first critical infrastructure initiative is the Critical Infrastructure Defense Program (CIDP). It gives trusted security providers access to frontier Claude models, on-site engineers, and threat research. The founding partners are:

• Accenture
• Booz Allen
• CrowdStrike
• Deloitte
• Dragos
• Hitachi
• Insane Cyber
• Nozomi Networks
• Palo Alto Networks
• PwC
• Rockwell Automation

Some partners are already using Claude to fix vulnerabilities. It's described as an early step, working first with a small group of providers to learn what works best.

The first open-source product is OSS Scanner, an opt-in service that gives enrolled projects periodic scans from Anthropic's strongest models at no cost.

• Each report includes a proof of concept, an explanation, and a suggested fix where possible
• Reports are sent without human review, so some may contain errors, such as incorrect severity ratings
• Anthropic expects a true-positive rate above 90% and plans to improve it
• Projects that can't keep up with the volume will continue to receive human-verified disclosures

Anthropic also funds open-source foundations and vulnerability-coordination groups, and supports free Claude Max subscriptions for maintainers.

Next, Anthropic plans to expand the CIDP to more partners and sectors, share lessons (including failures), collaborate with other AI developers, security firms, and governments, and extend its open-source work to the broader supply chain.

Wrap-up

• The Anthropic Cyber Mission is a long-term effort to support defenders
• The CIDP for critical infrastructure provides Claude models, on-site engineers, and threat research
• OSS Scanner provides free periodic scans from top-tier models to enrolled projects
• Reports arrive without human review, so they may contain errors such as wrong severity

If you maintain open source, work on OT and critical infrastructure security, or run a security firm, check out the interest forms and programs!