TL;DR
US control over the most cyber-capable AI models, led by Anthropic’s Claude Mythos, looms over the NATO summit in Ankara on 7-8 July. Washington has whipsawed between export controls and expanded allied access via Project Glasswing, frustrating European allies who are demanding access while building their own defence AI. Officially, the summit will barely mention it.
Donald Trump arrives at next week’s NATO summit in Ankara holding unusual leverage, because the US decides which allies get access to the world’s most advanced AI, Politico reports. The alliance meets on 7 and 8 July with AI security questions hovering over the agenda.
A new wave of models from Anthropic and OpenAI can find and exploit security flaws better than most human specialists. Anthropic’s Claude Mythos surfaced vulnerabilities in classified US systems within hours during a government test.
“AI is fundamentally changing the threat landscape, and NATO needs to adapt accordingly,” Estonian cyber ambassador Helen Popp told Politico. Every capability available to adversaries is also available to allies, she argued, if they move first.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
US agencies including the NSA and CISA have been testing Mythos for cyber defence and digital espionage. European allies have clamoured for access, and EU institutions have openly demanded it, with only a few countries, including the UK, initially allowed to run evaluations.
Anthropic expanded its Project Glasswing programme in June to around 150 organisations across more than 15 countries, including the EU. The scramble followed weeks of whiplash from Washington.
In early June, the Trump administration imposed export controls on Anthropic’s most cyber-capable models, banning foreign nationals from using them and forcing a worldwide shutdown. The controls were lifted on 30 June after an 18-day blackout.
The White House has also limited the rollout of OpenAI’s latest model to a small group of approved US firms, per Politico. The push and pull has frustrated allies, prompted a rare Five Eyes warning on AI cyber threats, and left frontier models moving between governments faster than regulators can follow.
Quiet corridors, loud subtext
The summit agenda includes a track on emerging and disruptive technologies, but an official told Politico that AI and cyber will get only brief mentions in the closing statement. Former NATO cyber policy leader Heli Tiirmaa-Klaar said allies avoid formally discussing topics that lack consensus, predicting talks in the margins instead.
The US State Department’s cyber bureau is not sending a representative amid an internal reorganisation, Politico reports. Senator Jeanne Shaheen said she will attend partly to reassure allies that the US will not “alienate them” over access to AI models.
Trump has separately signed NSPM-11, ordering the US military to adopt AI faster and shield models from China. Europe is hedging by building its own capability, including the defence AI alliance between Helsing and Mistral.
The war in Ukraine, now past its fourth year, keeps the stakes concrete, and allies have pledged 1.5% of GDP to protecting critical infrastructure. Laura Galante of the Center for European Policy Analysis called Ukraine the blueprint for operating in AI-fuelled warfare.
A State Department spokesperson said every ally must adopt “trusted leading-edge AI capabilities”. Which capabilities count as trusted, and who grants the trust, is precisely what Ankara will not quite discuss.
