THE CRUNCH
OpenAI has opened its cybersecurity tool Daybreak to NATO and several European governments, letting cyber defence teams put its models to work auditing software, hunting for weaknesses and closing security holes. The NATO Communications and Information Agency (NCIA) says it is already leaning on the underlying models heavily, and its chief Dylan Browne called the tooling a generational leap in cyber defence.
The models have already produced concrete results. ENISA, the EU's cybersecurity agency, used them to uncover four flaws in software that EU institutions rely on, while a joint report with CERT-EU looked at 27 software projects and found that 88 reported issues held up in 95 percent of cases. Poland's CERT Polska, meanwhile, identified six flaws in MikroTik routers that attackers were exploiting at the time.
Ukraine has put Daybreak to work defending civilian infrastructure, and France, Germany and Poland have also been granted access. At NATO's Crossed Swords 2026 exercise, participants for the first time ran AI agents for cyber operations with humans keeping oversight.
Earlier on this story
Crunchy Tech briefs on this story.


