The company will provide support to critical infrastructure operators and offer free vulnerability scanning services for open-source software.

Anthropic announced on Thursday the launch of a new program aimed at supporting cybersecurity for critical infrastructure and open-source software.

In a blog post, Anthropic stated that the new initiative, called the "Anthropic Cyber Mission," will offer threat research, on-site engineer services, and access to its frontier Claude large model to infrastructure operators. Such infrastructure operators span power grids, utilities, factories, and transportation networks.

In addition, Anthropic will provide regular free vulnerability scanning for open-source software projects, with remediation suggestions generated by the company's most capable models.

Just this Tuesday, Anthropic had announced that it would open access to its frontier large models for cyber defenders, allowing participants to use Claude models with regular guardrails removed (guardrails that would otherwise prevent the models from being used for cyberattacks). The "Cyber Mission" launched on this basis.

In recent months, multiple high-profile hacking incidents have involved AI models, both from malicious human actors and runaway AI agents, raising industry concerns about the cybersecurity risks posed by high-performance AI.