3 Sources Reveal G7 'Trusted Partner' Plan to Bypass US AI Export Ban

For legal professionals, the nascent G7 'trusted partner' scheme represents a critical test of how far US executive authority can stretch in extraterritorially controlling AI technology. The proposal, discussed with Commerce Secretary Howard Lutnick, could set a precedent for future technology embargoes, challenging existing WTO rules and bilateral trade agreements. Law firms and compliance officers must watch how this exception is legally structured—its scope, vetting criteria, and enforcement mechanisms—as it may redefine the boundaries of technology transfer regulations.

At the G7 summit in Evian-les-Bains, France, a pivotal discussion has emerged that could reshape the global landscape for advanced artificial intelligence. After President Donald Trump directed Anthropic to bar foreign nationals from its most sophisticated AI models—Fable 5 and Mythos 5, disabled for all users on June 13—diplomatic sources revealed that G7 leaders are exploring a 'trusted partner' scheme. This plan, discussed on the sidelines of the summit's opening dinner on June 16 with U.S. Commerce Secretary Howard Lutnick, would grant select allies or companies exemptions from the ban, effectively creating a privileged tier of access to cutting‑edge U.S. technology. The move underscores a critical tension: the United States' national security imperative to safeguard AI from adversaries like China versus the need to equip democratic allies with the same tools to maintain technological parity and fortify cybersecurity.

“

Moreover, the lack of a formal statement on the matter during the G7’s tech‑focused working lunch on June 18, where CEOs from Anthropic, OpenAI, and Google are expected, suggests that the concept is still in early diplomatic trial balloon stage.

The immediate catalyst for these talks is the abrupt cutoff Anthropic enforced after the White House order. Before the prohibition, Mythos 5 had been shared with select organizations in more than 15 countries, according to prior reports. The model’s capability to detect vulnerabilities in computer code is seen as a double‑edged sword: cybersecurity experts warn it could 'turbocharge attacks on banks’ technology systems,' yet it is also a vital instrument for defensive security research. The European Union has formally requested access to Mythos to study its implications, reflecting a broader anxiety among G7 members that being locked out of the most advanced AI will weaken their cyber defenses and competitive posture. By proposing a 'trusted partner' status, possibly for both nations and corporations, the G7 is seeking a policy bridge that maintains security while fostering collaboration against common threats.

The legal and regulatory dimensions are substantial. The current restriction echoes previous U.S. export control regimes—such as those applied to Huawei—but extends them to intangible AI models, a frontier where jurisdictional reach and enforcement are far murkier. A trusted‑partner exception could sidestep the absolute ban, but its design will be contentious. Who qualifies? What vetting mechanisms and ongoing compliance obligations would apply? For the European Union, which runs strict data protection and AI governance frameworks under the GDPR and the upcoming AI Act, integrating such an exception without undermining its regulatory autonomy will be a delicate diplomatic mission. Moreover, the lack of a formal statement on the matter during the G7’s tech‑focused working lunch on June 18, where CEOs from Anthropic, OpenAI, and Google are expected, suggests that the concept is still in early diplomatic trial balloon stage.