US frontier AI access looms over NATO Ankara summit, shaping alliance interoperability and defence AI development

US control over the most cyber-capable frontier AI models, including Anthropic's Claude Mythos, is looming over the NATO summit in Ankara on 7-8 July, shaping how allies can access and deploy these systems for defence. The summit's official agenda barely mentions AI, but the access dispute is driving European allies to accelerate their own defence AI initiatives and is creating new constraints for anyone building AI products for alliance use.

What happened

The United States decides which NATO allies can use the world's most advanced AI models. Anthropic's Claude Mythos, a frontier AI model capable of finding vulnerabilities in classified US systems within hours during government tests, is at the center of the dispute. In June, the Trump administration imposed export controls on Mythos, which were lifted on 30 June after an 18-day blackout. The same reporting notes that OpenAI's latest model rollout remains limited to a small set of approved US firms.

Anthropic expanded its Project Glasswing programme in June to around 150 organisations across more than 15 countries, including the EU. This expansion followed weeks of whiplash between restrictions and expanded allied access. European allies have demanded access, with only a few countries, including the UK, initially allowed to run evaluations. The tensions prompted a rare Five Eyes warning on AI cyber threats.

Europe is hedging by building its own capability, including the defence AI alliance between Helsing and Mistral. Meanwhile, the war in Ukraine has driven allies to pledge about 1.5% of GDP to protecting critical infrastructure, keeping the stakes concrete.

Why AI builders should care

Access governance of high-end AI models shapes interoperability standards, risk management, and supply chain decisions for defence AI deployments. If you build AI for NATO member states or critical infrastructure, you must account for who controls model access and under what terms. The dispute between US controls and allied ambitions may accelerate Europe's development of its own defence AI capabilities and governance frameworks. That could create parallel model ecosystems with different compliance requirements.

Practical implications

Developers building AI for defence or critical infrastructure must consider cross-border access constraints when designing interoperable systems. Policy signals from the Ankara margins, even if the official statement is brief, may meaningfully impact deployment choices. Expect EU and allied actors to pursue alternative routes to ensure access outside US-controlled channels, potentially through joint procurement or local model development initiatives.

Caveats

The article notes that the summit's official communique is unlikely to dwell on AI access, with discussions expected in the margins instead. All factual claims are drawn from the cited TNW article and may not reflect all perspectives. The US State Department's cyber bureau is not sending a representative amid an internal reorganisation, which may limit formal outcomes.