Anthropic announced the launch of a new cybersecurity initiative aimed at making the company’s most advanced AI models and its specialist engineers available to firms responsible for protecting power grids, water facilities, and other critical infrastructure. The news was first shared exclusively with Axios.
According to ITNA citing Axios, operators of critical infrastructure are currently facing a serious challenge securing their legacy and complex systems - precisely as AI can make executing cyberattacks faster and cheaper.
According to the report, Anthropic, under the “Critical Infrastructure Defense Program,” will provide its advanced AI models, engineers, and research findings on cyber threats to companies responsible for securing infrastructure systems.
These companies are expected to use these resources to identify and remediate vulnerabilities in their customers’ systems.
Founding partners of the program include Accenture, Booz Allen Hamilton, CrowdStrike, Deloitte, Dragos, Hitachi, InSync Cyber, Nozomi Networks, Palo Alto Networks, PDC, and Rockwell Automation.
Anthropic says some of these partners are already using cloud models to remediate vulnerabilities and assist their customers in this area.
Free Vulnerability Scanning Service
Anthropic also announced the launch of a free AI-based vulnerability scanning service for open-source projects called OSS Scanner.
According to the report, participating projects will be periodically scanned by the cloud and will receive automated reports about potential vulnerabilities, how they could be exploited, and suggested remediation steps.
However, no human review is performed before these reports are sent to developers; Anthropic warns this may lead to some incorrect findings in the reports.
Anthropic, OpenAI, and other AI developers are racing to put their most powerful models in the hands of cyber defenders as hackers gain access to similar capabilities.
Anthropic’s new program builds on experience from Project Glasswing, which previously gave approved organizations access to the company’s most powerful models to identify security vulnerabilities. Anthropic says that project demonstrated how quickly AI can discover vulnerabilities, while verification and remediation remain difficult tasks.
It remains unclear how partner companies will test and deploy security solutions without disrupting infrastructure operations - one of the biggest challenges in critical infrastructure security. Anthropic also has not specified whether partners will get free access to the models or who will bear the computational costs of using the company’s AI tools.