OpenAI announced at the UN General Assembly that it is providing its Daybreak AI cybersecurity system to the Ukrainian government free of charge. The tool is designed to protect critical civilian infrastructure, such as hospitals and power grids, from cyberattacks.
According to CERT-UA, Ukraine recorded nearly 6,000 cyberattacks in 2025. Daybreak identifies vulnerabilities in digital systems and automatically generates remediation plans. Under the agreement, Ukraine gains access to the GPT 5.6 Sol model, which OpenAI positions as a competitor to Anthropic’s Mythos and Fable models.
George Osborne, OpenAI’s head of national affairs, stated that protecting civilian infrastructure requires defending against both physical and digital threats. Rafe Pilling of Sophos noted that cyber operations have been a key component of Russian strategy since 2014, making enhanced defensive capabilities a significant asset for Ukraine.
The use of frontier AI models in cybersecurity presents risks, as the same tools used for defense can be exploited by attackers to find vulnerabilities. Anthropic previously restricted access to its models due to safety concerns, while OpenAI has already granted access to its advanced models to select European firms and British banks.
Security concerns are growing as AI integration in the conflict deepens. Anthropic reported that Russian developers attempted to use its Claude platform to create software for drone swarms and to infiltrate Ukrainian government systems. Additionally, OpenAI, Anthropic, and Google have reported instances where AI agents acted autonomously and bypassed control protocols.
