Anthropic Cyber Mission is now available

Anthropic launched the Anthropic Cyber Mission, an initiative deploying frontier AI models, threat research, and engineering talent to defend operational systems and public codebases, according to Anthropic News's report. The effort focuses on two critical areas: protecting operational technology across utilities and public transport, and scanning open-source software for software vulnerabilities. As state-sponsored adversaries seek footholds across industrial control systems and enterprise supply chains, the initiative brings frontier Claude models to operational technology defenders and open-source maintainers facing ongoing resource shortages.

This operational briefing examines the architecture of the initiative, the industrial defense program, and the automated tools distributed to software maintainers. The program targets operational technology engineers, industrial control system operators, and open-source maintainers managing legacy systems that cannot be taken offline for routine maintenance. By pairing frontier model access with technical personnel from Anthropic, the initiative establishes a direct channel for identifying vulnerabilities, prioritizing exposures, and generating functional remediations for high-consequence environments.

What shipped in Anthropic Cyber Mission is now available

Anthropic rolled out two central operational mechanisms under the Cyber Mission initiative: the Critical Infrastructure Defense Program and the OSS Scanner tool. The Critical Infrastructure Defense Program supplies frontier Claude models, threat intelligence, and on-site Anthropic engineers directly to security providers and industrial manufacturers. Alongside this, Anthropic launched OSS Scanner, a utility that provides open-source software projects with regular, automated security scans powered by the company's strongest models at no cost. These offerings address the bottleneck of parsing complex systems without introducing downtime to sensitive operations.

The release also incorporates the merger of Project Glasswing into an expanded Cyber Verification Program, widening access to frontier models for external defenders. In operational environments, power grids, water networks, factories, and transportation grids operate on legacy controllers and industrial networks engineered to run undisturbed for decades. The newly launched programs deliver technical evaluation tooling and verified guidance directly to the technical partners who maintain and patch these industrial systems.

What improved in Anthropic Cyber Mission is now available

The transition from earlier exploratory vulnerability initiatives to the expanded Cyber Verification Program alters how defenders isolate, verify, and remediate discovered exposures. Earlier testing under Project Glasswing confirmed that while discovering vulnerabilities had become straightforward, prioritizing and safely remediating flaws remained difficult. The new framework pairs automated model discovery with human domain expertise and on-site engineering support to accelerate remediation cycles.

The program incorporates eleven founding industrial and cybersecurity partners: Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC, and Rockwell Automation. This coalition combines Palo Alto Networks Unit 42 threat intelligence, industrial device context from Rockwell Automation, and CrowdStrike endpoint telemetry with Claude models. Integrating frontier model analysis directly into partner response workflows reduces the time needed to evaluate high-volume telemetry across remote substations and air-gapped industrial facilities.

What you gain from Anthropic Cyber Mission is now available

Industrial operators and critical software maintainers gain specialized analysis capabilities capable of processing operational data without requiring plants to suspend continuous physical processes. Because operational technology cannot easily be taken offline for patching, operators rely on trusted partners to assess exposures and evaluate mitigations in place. Through the Critical Infrastructure Defense Program, operators receive threat evaluations backed by Anthropic technical teams who assist partner organizations in structuring safe remediations.

Open-source maintainers gain direct access to regular security reviews from top-tier models without incurring commercial API expenses or maintaining dedicated security pipelines. For specialized environments like offshore energy platforms and isolated substations where security personnel cannot be continuously stationed, automated analysis processes massive telemetry volumes so human engineers can focus exclusively on critical remediation decisions. This shared defensive posture allows operators to close exposure paths before adversaries exploit them.

How to get Anthropic Cyber Mission is now available

Organizations providing security products or engineering services to critical infrastructure sectors can register their interest directly through Anthropic's partner portal. Maintainers and infrastructure engineers integrating Claude into local defense environments can update their development tools and configure defensive model workflows through the command-line interface.

npm install -g @anthropic-ai/claude-code

Security teams switching between model tiers during vulnerability triage and verification sessions can invoke the operational model interactively within the development environment:

/model claude-fable-5-1

To execute pinned model sessions directly from the terminal for unattended vulnerability verification pipelines, operators launch the environment with the target model flag:

claude --model fable

Defenders seeking to configure permanent environment defaults for enterprise analysis scripts can set the model target in their local shell environment configuration:

export ANTHROPIC_MODEL="claude-fable-5-1"

What to watch after Anthropic Cyber Mission is now available

Defenders should monitor initial performance benchmarks and operational reports emerging from the founding cohort of eleven industry partners. Because industrial control systems and operational technology encompass proprietary protocols and legacy hardware, the practical efficacy of model-assisted patching will depend on validation data generated across power utilities, manufacturing plants, and municipal water facilities. Observations gathered from this initial cohort will dictate how Anthropic expands the program across additional infrastructure sectors.

The open-source community will also monitor the rollout of OSS Scanner to assess patch quality and verification accuracy across public repositories. Anthropic confirmed plans to expand the Cyber Mission with additional tools, research publications, and defense funding over time. Defenders should watch how automated scanning tools balance detection volume with false-positive filtering, ensuring that open-source maintainers receive actionable, verified pull requests rather than unvalidated vulnerability reports.