$1B in subsidized models for water and power utilities

OpenAI president Greg Brockman on September 3 committed $1 billion to Daybreak for Frontline Defenders, a program that gives water systems, electricity providers, local governments and other critical services subsidized access to OpenAI models. He announced it while hosting 300 security leaders at the company’s headquarters, Axios reported.

The $1 billion is paid out in kind: expanded subsidized model access, training, technical support and partnerships for participating organizations, aimed at essential services around the world. A companion program, Daybreak for America, covers local governments, water systems, electricity providers, regional banks and other critical infrastructure. It starts with a pilot run with the Multi-State Information Sharing and Analysis Center, which will train cyber defenders at state and local governments to use the models.

OpenAI lists the work it expects lesser-resourced utilities to do with them: review legacy code, analyze suspicious activity, identify and validate vulnerabilities, prioritize the most serious risks, and develop and test security patches. More than 35 technology and cybersecurity services are also embedding OpenAI’s advanced models into their tools, services and workflows.

Daybreak itself is older than this money. OpenAI introduced it in May, shortly after Anthropic launched its own cybersecurity coalition, Project Glasswing, which Anthropic expanded in June to 150 new organizations across 15 countries, energy and water among them.

On August 10 OpenAI split Daybreak into two tiers. Daybreak Blue offers advanced general-purpose models with safeguards altered to allow defensive security work. Daybreak Red adds purpose-trained cybersecurity models for security testing, vulnerability research and exploit validation, including GPT-5.6-Cyber, built on GPT-5.6 Sol and tuned to refuse fewer specialized security tasks. OpenAI recommends Blue as the starting point for most organizations, CNBC reported.

The pressure behind the program comes from both sides of the model market. Cyber officials have long feared that AI models will soon be able to find and chain together vulnerabilities in the systems that run public utilities, autonomously. Frontier labs, OpenAI among them, have largely restricted the advanced cyber capabilities of their most powerful models. But open-weight models are improving quickly, and that is raising fears that AI-enabled attacks will arrive sooner than expected, at organizations that have long lacked the budget, staff and time to harden their defenses.

Axios adds that more services and training won’t fix the security problems critical infrastructure has carried for decades, which AI-enabled tools will likely make worse. Many of these operators also need more workers who understand how to protect the physical systems they run from digital threats.