OpenAI is launching a program to provide AI models in the form of subsidies to critical infrastructure organisations such as water utilities, power suppliers and local governments, Axios reported on Sept. 3 (local time).

According to the report, OpenAI President Greg Brockman (그렉 브록먼) announced the new program to 300 security officials at a security summit held at the company’s headquarters.

OpenAI will first invest $1 billion in the Daybreak for Frontline Defenders initiative as part of efforts to protect critical services worldwide.

The funding will support expanded model access, training, technical support and partnerships for participating organisations. OpenAI is also launching the Daybreak for America program. The programme centres on supporting model access for critical infrastructure organisations such as local governments, water utilities, power suppliers and regional banks.

OpenAI will also work with MS-ISAC (Multi-State Information Sharing and Analysis Center) on a pilot program to train state and local government cyber defence officials on how to use OpenAI models. OpenAI will also work with more than 35 technology and cybersecurity companies. The focus is on embedding OpenAI’s advanced models into these companies’ tools, services and work processes.

Security officials have raised concerns that AI models could gain the ability to independently find and connect vulnerabilities in complex systems that run public facilities and services. Leading AI companies including OpenAI have generally restricted advanced cyber capabilities in their top models, but concerns are growing that AI-based attacks could become reality sooner than expected as open-source model performance improves rapidly.

OpenAI explained that the program will enable cash-strapped facility operators to review legacy code, analyse suspicious activity, identify and validate vulnerabilities, set risk priorities, and develop and test security patches.

Some have pointed out that expanded services and training alone will not resolve infrastructure security problems that have persisted for decades. There are also concerns that AI-based tools could worsen the problem, and many organisations also need additional personnel with knowledge to protect physical systems from digital threats, Axios reported.