OpenAI gives Ukraine its cyber defence tool⁠.

OpenAI is extending access to its Daybreak programme to the Government of Ukraine, to support the cyber defence of civilian infrastructure.¹ The BBC reports that the system will be shared for free, to help protect hospitals and power plants from cyber attacks.² Daybreak can find weaknesses in digital systems quickly and help develop fixes.² Under the deal Ukraine also gets access to OpenAI’s GPT 5.6 Sol model, which the BBC describes as a rival to Anthropic’s Mythos and Fable.²

Ukraine’s national cyber incident response team, CERT-UA, recorded nearly 6,000 attacks in 2025.² George Osborne, the former UK chancellor who now heads OpenAI for Countries, said civilian infrastructure has to be defended against both physical and digital attacks so that people can keep living, working and using essential services.² Rafe Pilling of the security firm Sophos said Russian offensive cyber operations had been a key component of the Kremlin’s aggression against Ukraine since 2014.²

Jamie MacColl, a senior research fellow at the Royal United Services Institute, told the BBC that western tech firms have backed Ukraine partly for altruistic reasons and partly because an active conflict gives them very valuable data and intelligence.² He said Daybreak would be useful given the volume of Russian attacks, but noted that Ukraine already has AI tools from OpenAI’s competitors, Google among them.²

The skill that lets a model spot weaknesses for a defender is the same one an attacker uses to hunt for a way in.² Anthropic has tightly restricted access to its most powerful systems on the grounds that they could fall into the wrong hands, while OpenAI has let some firms in Europe, and UK banks, use its most advanced cyber models.² Earlier this month Anthropic reported that its Claude platform appeared to have been used by a group of Russian developers to build software for attack drones, after they used VPNs to get round a geographic block.²

The announcement came alongside the UN General Assembly in New York, where AI is firmly on the agenda.² Around 100 US companies recently signed an open letter warning that the window is closing for cyber defences to be ready for AI enabled attacks.²

What it means for your defences

For a UK company the direct effect is nil. This is a deal between a lab and a government under attack. The detail worth noticing is the one about UK banks. The strongest cyber models are being handed out selectively, to governments and large regulated firms first, and a company of fifty or five hundred staff is not near the front of that queue.

So the sensible move is the unglamorous one. For most firms the weaknesses these systems find are the old ones: software nobody patched, passwords used twice, a supplier with more access than it needs. You do not need Daybreak to close those. You need a list, an owner and a date. If a supplier offers you AI driven security, ask what it found last month and what it fixed, not which model it runs.

Also today

• Dr Martens uses Salesforce agents to win back customers

  Dr Martens says AI agents, built with Salesforce as part of a wider digital overhaul, are helping to turn round customer satisfaction, which had been slipping.⁶

• Agentic AI needs its rules before it goes live

  Computer Weekly argues that IT leaders racing to deliver agentic AI risk setting controls only after proving the concept, and that governance for agents has to be in place from the start.³

• Pornhub investigated over how it checks users’ ages

  Pornhub is under investigation over its age checks, with the regulator concerned about how the site relies on third party checks provided by Apple for some of its users.⁵

• UKHSA extends wastewater testing for dangerous pathogens

  UKHSA labs will extend wastewater surveillance to hospital sites and busy public locations, paid for with £3 million from the UK Integrated Security Fund, as an early warning system.⁴