OpenAI offers Ukraine its cyber programme, after CERT Polska found six flaws with its models
OpenAI says it will offer Ukraine's government access to Daybreak, its restricted programme for authorised security work, and the clearest result that programme has produced so far belongs to Poland. Poland's national cyber agency CERT Polska used OpenAI models to help find six vulnerabilities in third-party router software, according to OpenAI's announcement of September 23, and the vendor has since released fixes.

OpenAI's title card for Sam Altman's remarks at the UN Security Council, published the same day, September 23. Image: OpenAI
CERT Polska did the finding. OpenAI's wording is that the agency "used OpenAI models to help discover six vulnerabilities", and that the fixes the vendor shipped are ones "CERT Polska confirms prevent the attacks it observed". Neither the router vendor nor the model is named.
An offer, not a deployment
What Ukraine gets is access. Working with Ukraine's Ministry of Digital Transformation, the company says it "will provide Ukrainian teams with access to tools to identify software vulnerabilities and develop and test fixes more quickly".
Daybreak itself is not a product anyone can sign up for. OpenAI describes it as access to advanced AI "for authorized security work, from reviewing older software and investigating suspicious activity to validating vulnerabilities and testing fixes", which is the same gated footing as the cyber model OpenAI announced in August without opening it to the public.
The announcement was made on the sidelines of the UN General Assembly by Dmytro Kushneruk, Ukraine's Consul General in San Francisco, and Sasha Baker, OpenAI's Head of National Security Policy. "Ukraine is already on the front line, and its defenders need support now," Baker says. "We want to put more capable tools in their hands to help them find and fix vulnerabilities and protect the critical networks people depend on."
The other name on the page is George Osborne, listed as Head of OpenAI for Countries. "Protecting civilian infrastructure means defending it against both physical and digital attacks, so people can continue to live, work and access essential services," he says.
CERT-UA logged nearly 6,000 incidents in 2025
OpenAI puts the scale of the problem at nearly 6,000 cyber incidents handled by CERT-UA, Ukraine's national incident response team, during 2025. Those include attacks on hospital systems, the energy sector and telecommunications.
Ukraine is not the first government on the list. OpenAI says it has provided cyber model access to defenders in France, Germany, Poland "and others", and that the EU's cyber agency ENISA has used the models to identify vulnerabilities in software used across EU institutions, "all of which have since been fixed". No count is given for ENISA's findings, which leaves CERT Polska's six as the only number OpenAI puts a name to.
OpenAI is not the only lab handing security-capable models to defenders and then publishing the result. Google said in September that Gemini broke into three real companies during a May security test.
For Ukraine, the announcement stops at the offer. OpenAI has not said which model Ukrainian teams will get, when the access begins, or what limits come with it.