Anthropic launches Cyber Mission to defend power grids and open-source code
The AI lab is pairing frontier models, funding and a roster of security heavyweights to help defenders protect critical infrastructure and open-source software
Anthropic has launched the Anthropic Cyber Mission, a new initiative aimed at protecting critical infrastructure and open-source software. The tools for the job are frontier AI models, engineering resources, threat research and money.
The launch date was October 8, 2026. The timing reflects a shift in the threat landscape: AI is increasingly being used to attack systems, so defenders want the same firepower on their side.
Two fronts: the grid and the code
The mission concentrates on two areas. The first is operational technology, or OT, which covers power grids, water systems and transportation networks.
That is also what makes OT hard to secure. Many of these systems cannot be taken offline for patching, which means the process of verifying and fixing a vulnerability is far more burdensome than updating a laptop.
The second front is open-source software, or OSS. This is the freely shared code that underpins a huge share of the world’s applications, often maintained by small teams of volunteers.
Here, the headline tool is the OSS Scanner. It offers automated vulnerability assessments to maintainers of eligible open-source projects, with a projected true-positive rate exceeding 90%.
According to the research findings, the scanner delivers support to maintainers without prior human review. That speeds things up for maintainers who rarely have spare hours to triage reports.
The money and the partners
Anthropic is backing the mission with real resources. The company has committed $100 million in usage credits through Project Glasswing.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
It has also donated $4 million to various organizations. Recipients include OpenSSF and the Apache Software Foundation.
On the infrastructure side, the Critical Infrastructure Defense Program has 11 founding partners. Among them are Accenture, Deloitte and CrowdStrike, with Palo Alto Networks also named in the partner group.
The program draws on Anthropic’s most advanced models, including Claude Mythos. Access comes with responsible controls designed to keep the tools from being turned against the systems they are meant to protect.
Built on Project Glasswing
The Cyber Mission did not appear out of nowhere. It builds on Project Glasswing, which Anthropic started in April 2026.
Glasswing has since been folded into the company’s expanded Cyber Verification Program. In effect, a pilot project has graduated into the foundation of a much larger operation.
Software vulnerabilities are multiplying, and they are increasingly easy to find. The bottleneck is no longer spotting flaws. It is verifying and fixing them before someone else exploits them.
What this means
For open-source maintainers, a scanner with a projected true-positive rate above 90% could take real weight off overstretched volunteers, provided that projection holds up in practice across varied codebases.
The $4 million in donations to groups like OpenSSF and the Apache Software Foundation also matters. Open-source security has long suffered from a funding mismatch: enormous economic value resting on very thin resources.
The access controls are the piece to watch. Tools strong enough to find serious vulnerabilities are, by definition, tools an attacker would love to have.
Anthropic is betting that the same technology fueling offensive cyber operations can tilt the balance back toward defenders. With $100 million in credits, 11 founding infrastructure partners and a scanner aimed at the open-source commons, the company has put meaningful resources behind that bet.