Anthropic is starting the Anthropic Cyber Mission, a long-term commitment to help secure the systems everyone relies on. The effort gives defenders tools, research, and resources to protect their software and systems. It begins in two areas:
- Critical infrastructure: The first focus is the operational technology behind power grids, water systems, and transportation networks, along with government systems. The new Critical Infrastructure Defense Program (CIDP) gives the defenders of operational technology frontier models, on-site engineers, and threat research.
- Open-source software: Anthropic will find vulnerabilities, propose patches, and harden the shared code that most software depends on. It is launching OSS Scanner, which gives open-source projects regular security scans from Anthropic's strongest models at no cost.
Frontier models can be misused to exploit vulnerabilities and run cyber operations, and many areas of technology remain dangerously exposed. State-sponsored adversaries have spent years gaining footholds in systems across many sectors so they can disrupt them later. Defenders of critical infrastructure and the open-source community have decades of experience but face severe resource shortages, which the current moment makes worse. Anthropic recognizes their expertise and is offering support. It plans to extend the Cyber Mission with tools, research, and resources in new areas.
Partners in Project Glasswing found many vulnerabilities, but cyber risk has not yet dropped enough. Finding vulnerabilities is easier than ever, while verifying, prioritizing, and fixing them is still hard. Using lessons from Glasswing, Anthropic merged the project into its expanded Cyber Verification Program earlier this week, giving many more defenders access to Anthropic's most capable models. Through the Cyber Mission, Anthropic will also deploy engineering talent and provide tools and funding for those securing critical infrastructure and open-source software.
Defending critical infrastructure
Power grids, water utilities, factories, and transportation networks run on controllers, control software, and industrial networks designed to last for decades. This is called operational technology. It often can't be taken offline for patching, so known vulnerabilities can stay open for years. The work is specialized: equipment is proprietary, changes are risky, and one mistake can shut down a plant. Operators of every size depend on a small set of trusted providers to tell them what is exposed and which fixes to apply to a running system.
The Critical Infrastructure Defense Program gives those trusted providers frontier Claude models, on-site engineers, and Anthropic's threat research. The founding partners are Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC, and Rockwell Automation.
These organizations represent the providers operators rely on: consulting and advanced technology firms that run security programs, security companies that protect business and industrial networks, and manufacturers that build and patch the equipment.
The work is already underway. Several partners are using Claude to fix vulnerabilities and help their customers do likewise. AI can't solve many of the difficulties of defending critical infrastructure. Anthropic still believes frontier models can help find and repair weaknesses before they are used to cut power or make water unsafe. The first step is working with a small group of providers to learn which strategies are most effective and practical.
Companies that build security products or services for critical infrastructure can register their interest in the program.
Partner statements:
- Tony Baker, Rockwell Automation: AI offers an important chance to strengthen critical infrastructure cybersecurity if it is applied responsibly, validated rigorously, and deployed with safety and reliability first. Rockwell is glad to contribute its OT and industrial cybersecurity expertise to practical approaches.
- Harshul Joshi, PwC: PwC's cyber OT practice has seen the serious consequences of breaches in these environments. He believes a coordinated, industry-wide effort is needed to identify vulnerabilities, reduce risk, and build resilience.
- Sam Rubin, Palo Alto Networks: Threat actors use AI to find and exploit exposures at machine speed. Combining Unit 42 intelligence with Anthropic's models gives defenders the speed and precision to close attack paths before exposure becomes impact.
- Moreno Carullo, Nozomi Networks: Frontier AI is evolving faster than any one organization can follow. The program joins people who understand these environments with frontier capabilities, and Nozomi is proud to contribute and learn.
- Dan Gunter, Insane Cyber: The sites needing protection most, such as remote substations, offshore platforms, and air-gapped plants, are where traditional tools and specialists can't easily reach. Claude helps close the gap in data volume and site coverage so analysts and operators can focus on key judgment calls.
- Jon Lavender, Dragos: Critical infrastructure benefits when vendors share what they learn about AI in operational environments instead of keeping it siloed. Joining the program helps align frontier AI with a collective defense strategy built around real OT.
- Daniel Bernard, CrowdStrike: Critical infrastructure is where cyber risk becomes real-world risk. Machine-speed threats require machine-speed defense, and pairing CrowdStrike's security context with Anthropic's AI expertise gives defenders a greater advantage.
- Hiroaki Koiwa, Hitachi: With AI and cyber threats advancing quickly, securing OT environments that support social infrastructure is among society's highest priorities. Hitachi looks forward to collaborating to improve resilience.
- Adnan Amjad, Deloitte: As AI accelerates, operators need solutions that address vulnerabilities at machine speed. Deloitte and Anthropic are helping enterprises close security gaps and strengthen resilience.
- Andrew Turner, Booz Allen: OT is the next frontier for autonomous AI-enabled attacks. The question is how much control AI can gain and how fast. The program convenes private-sector innovation and investment to strengthen OT and ICS resilience.
- Harpreet Sidhu, Accenture: Organizations that keep power on, medicine safe, food secure, and communications running face some of the most consequential cyber risks. Accenture will bring its OT security expertise to help put AI-powered defense in the right hands.
In June, Anthropic launched a cyber defense program for state, local, tribal, and territorial governments. Since then, it has offered frontier Claude models and technical support to more than half of US states and some of the country's largest public critical infrastructure operators. The support speeds up code scanning and patching, incident response, red teaming, and other security workflows.
Anthropic is listening to the people who know this landscape best. Its approach is to work with the companies, coalitions, governments, and nonprofits that operators trust, and to provide support where it is most needed.
Securing open-source software
Nearly all software relies on open-source code, much of it maintained by small volunteer teams. Under Project Glasswing, Anthropic scanned hundreds of widely used open-source projects, had humans triage and review many of the potential vulnerabilities, and privately reported findings to maintainers through its coordinated vulnerability disclosure process.
Some maintainers able to triage at scale asked for everything the models found in their projects, reviewed or not. In response, Anthropic is launching OSS Scanner, an opt-in service inspired by Google's OSS-Fuzz. Enrolled projects get periodic scans from Anthropic's most capable models for free. Each report includes a proof of concept showing how the bug could be exploited, an explanation, and a suggested fix where available. Reports are model-generated and sent without human review, so maintainers get them faster, but some will contain inaccuracies such as a wrong severity rating. Anthropic expects a true-positive rate above 90% and will work to improve it and fix quality over time.
The service suits projects with the capacity to keep up with the findings. For others, Anthropic will keep sharing human-verified disclosures under its coordinated disclosure policy.
OSS Scanner is a first step. Anthropic's open-source work aims to:
- Get findings to maintainers faster. Bring OSS Scanner to more projects while keeping human-verified disclosure for those that need it.
- Accelerate fixes. Automate triage and patching, which are still mostly manual, for projects that want it.
- Explore new secure architectures and coding practices. Research and share methods for hardening or rewriting code with projects that want to go beyond patching.
This will follow guidance from open-source maintainers and foundations, and Anthropic will share what works so others can replicate it. Anthropic has funded organizations behind widely used open-source code, including the Python Software Foundation, Alpha-Omega and OpenSSF through the Linux Foundation, and the Apache Software Foundation. It also supports Akrites and Gold Eagle, which collect and coordinate vulnerability reports from many sources so maintainers aren't overwhelmed. The Defender Advantage Fund (0xDAF), launched in August, supports pilot programs in these areas and keeps OSS Scanner free.
Maintainers can also apply to Claude for Open Source for free Claude Max subscriptions, and to the Cyber Verification Program for expanded access to Claude's cyber capabilities for defensive work.
Why Anthropic is doing this
Highly cyber-capable AI models are now widely available to attackers, but defensive tools, Anthropic's included, haven't reached enough of the defenders who need them. OSS Scanner, the Cyber Verification Program, defensive products partners build on Anthropic's platform, and Claude Security are some of the efforts here, with more to come as soon as possible.
Anthropic forecasts that within two years AI will favor defense: bugs will be easier to catch before release, software can be written to be secure from the start, and models can actively defend systems. In the near term, that may not hold. The cost of exploiting vulnerabilities has dropped, while verifying, disclosing, and fixing them is slow and still depends on people. In Glasswing, months often passed between finding and fixing a vulnerability. With operational technology, a fix may need to wait until it can be applied safely to running machinery, which in rare cases could take decades.
Success means critical infrastructure such as water, power, transport, and communications keeps running despite attacks by capable adversaries, with fewer exploitable entry paths and faster recovery. This will take years, and Anthropic expects the Cyber Mission to evolve as it learns, together with the public and private sectors, what works.
The path ahead
Over the coming months, Anthropic will bring the Critical Infrastructure Defense Program to more partners and sectors and share what it learns, including what didn't work. It expects other AI developers, security companies, and governments to run their own efforts and wants to collaborate with them wherever that helps defenders.
Anthropic will also expand its work to secure open source and the broader supply chain, including researching new ways of writing software and defending systems. It intends to share frontier research, tools, and resources to help others secure their software and supply chains.
Ways to work with Anthropic:
- Companies building security products or services for critical infrastructure, including security vendors, system integrators, and equipment manufacturers, can register interest in the Critical Infrastructure Defense Program.
- Core maintainers of critical open-source projects can enroll in OSS Scanner.
- Security teams of any size, including critical infrastructure operators and open-source maintainers, can apply to the expanded Cyber Verification Program for access to Anthropic's models.