OpenAI opens its Daybreak cybersecurity tool to NATO and European governments

OpenAI is giving NATO and several European governments access to its cybersecurity tool Daybreak. Through Daybreak, cyber defense teams can use OpenAI's models to audit software, find vulnerabilities and patch security gaps.
The NATO Communications and Information Agency (NCIA) says it is already using the underlying models heavily. NCIA chief Dylan Browne called it a generational leap in cyber defense; the source reports this in its own words, not as a direct quotation. During the NATO exercise Crossed Swords 2026, participants deployed AI agents for cyber operations under human oversight for the first time.
The source also lists results the models have already produced elsewhere. The EU's cybersecurity agency ENISA used them to find four vulnerabilities in software used by EU institutions. A joint report with CERT-EU examined 27 software projects, and 95 percent of the 88 reported issues turned out to be real bugs. Poland's CERT Polska found six vulnerabilities in MikroTik routers that were being actively exploited. Ukraine is using Daybreak to protect civilian infrastructure.
France, Germany and Poland have also been granted access. Of the "several" European governments, only these three are named.
Key facts
- OpenAI is giving NATO and several European governments access to Daybreak, a tool that lets cyber defense teams use its models to audit software, find vulnerabilities and patch security gaps.
- France, Germany and Poland have been granted access; Ukraine is using Daybreak to protect civilian infrastructure.
- ENISA used the models to find four vulnerabilities in software used by EU institutions; a joint ENISA and CERT-EU report examined 27 software projects, and 95 percent of the 88 reported issues turned out to be real bugs.
- CERT Polska found six vulnerabilities in MikroTik routers that were being actively exploited.
- At NATO's Crossed Swords 2026 exercise, participants deployed AI agents for cyber operations under human oversight for the first time.
Why it matters
The story is an expansion of access rather than a new capability: OpenAI's models are now offered to NATO and European governments for defensive cyber work through Daybreak. What gives it weight are the results the source cites. Government security teams report real findings, including actively exploited router flaws and a joint report in which 95 percent of 88 reported issues were genuine bugs. NATO's exercise also marked the first time participants deployed AI agents for cyber operations under human oversight.
Who it affects
NATO, through the NCIA, and European governments, with France, Germany and Poland named as having been granted access. Cyber defense teams are the intended users. EU institutions are affected as owners of the software ENISA and CERT-EU examined, and Ukraine uses Daybreak to protect civilian infrastructure.
How to use it
Daybreak is aimed at cyber defense teams, who can use OpenAI's models to audit software, find vulnerabilities and patch security gaps. No terms, pricing or conditions of access are given in the source. For most readers this is a signal about where government cyber defense is heading, not something to sign up for.
How solid is it
The source is a short news report from The Decoder. The figures come from named bodies: ENISA, CERT-EU, CERT Polska and the NCIA. The 95 percent figure is a share of the 88 reported issues in the joint report, not a share of projects. The source does not say whether it covers ENISA's four vulnerabilities or the whole report, and it gives no breakdown. Dylan Browne's remark is reported in the source's own words, not as a direct quotation.
Risks and caveats
The source gives no date for the announcement or for when access begins. It does not say which OpenAI models underlie Daybreak, or what the tool is technically beyond auditing, finding vulnerabilities and patching. It gives no details of the Crossed Swords 2026 exercise, and it names neither the MikroTik vulnerabilities nor the EU software affected. It mentions no criticism, security concerns or reactions from other parties.