One of the prominent topics at the G7 summit was access to U.S. advanced artificial intelligence models: participants discussed a proposal under which selected "trusted partners" could be exempted from recently introduced U.S. restrictions.
According to three diplomatic sources, G7 leaders considered allowing certain countries or companies limited access to the most advanced AI models developed by U.S. technology firms, such as those from Anthropic. The sources said the group of trusted partners could include both states and private companies.
Several delegations held talks on the idea with U.S. officials, notably with Howard Lutnick, the commerce minister, during background conversations surrounding the G7 opening dinner in Évian‑les‑Bains.
Last week Washington suspended foreign access to some of Anthropic’s top models on national security grounds. A focal point of the restrictions is Anthropic’s Mythos model, which was developed to uncover security vulnerabilities in software code. Cybersecurity experts have warned that Mythos could significantly increase the effectiveness of attacks against financial institutions’ IT systems.
To date, no European bank has received the Mythos model; the European Union has requested access in order to study the model’s potential impacts.
One source confirmed that participants did not expect a joint statement on the matter by Wednesday, the day when technology topics were to be formally placed on the agenda. The story was first reported by the Financial Times and the account has been corroborated by Reuters.
Why this matters
The discussions underline the tension between technological advancement, national security, and economic cooperation. The fact that G7 leaders considered limited access for selected partners suggests allied countries are seeking a balance between enabling innovation and managing security risks.
(This article is based on available reporting; the Financial Times first flagged the story and Reuters provided additional confirmation.)
