The G7 summit in Evian-les-Bains closed on June 17, 2026 with what the French presidency described as a dedicated AI working lunch between G7 political leaders and a selected group of technology executives — including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and representatives from Google DeepMind — to discuss ensuring safe, rapid and effective AI deployment. The working lunch format — heads of state and AI lab CEOs at the same table, specifically convened around AI — is a political signal that belongs in every enterprise AI governance discussion. The technology that enterprise organisations are deploying in production workflows, customer communication and financial decision-making has reached the level of international political attention that historically preceded the most consequential regulatory shifts in enterprise technology history.

President Trump said from the G7 that negotiations around the Fable 5 export control directive were "going fine." Anthropic's Managing Director of International Chris Ciauri said at the Seoul press conference on June 17-18 that he was "very confident" the models would return "within days." No official restoration date has been confirmed as of June 17.

The WIRED account of the Fable 5 trigger, published June 17, is the most important piece of context for understanding the government-Anthropic conflict and its enterprise implications. The White House identified SK Telecom — South Korea's largest carrier, a $100 million Anthropic investor since 2023, and a Project Glasswing partner — as a company suspected of having ties to China among the approximately 150 organisations granted Mythos access. The administration asked Anthropic to revoke only SK Telecom's access, which Anthropic immediately did. Separately, Amazon researchers then identified potential vulnerabilities in Fable 5 and reported them to the White House. The combined sequence — SK Telecom access revocation plus the Amazon vulnerability report — led the administration to conclude it could not trust Anthropic to safeguard its most advanced AI technology, triggering the broader export control directive.

David Sacks, co-chair of the President's Council of Advisers on Science and Technology, revealed on June 13 that the Trump administration offered Anthropic a choice before issuing the export control ban: Anthropic could either fix the identified jailbreak in Fable 5, or voluntarily de-deploy the model. Anthropic's refusal to voluntarily de-deploy — combined with its technical assessment that the vulnerability did not represent a meaningful capability advance beyond what other public models already offered — produced the export control directive that shut the model down for all customers.

The SK Telecom dimension of the story is the one that most directly affects enterprise AI vendor risk assessment. SK Telecom is not a marginal partner — it is a $100 million strategic investor in Anthropic and a member of Project Glasswing, the AI cybersecurity coalition we covered in April. The White House's designation of SK Telecom as a suspected China-linked entity while it was a Glasswing partner produced the chain of events that led to the export control directive. For enterprise AI procurement teams, the implication is specific: an AI vendor's strategic partnerships — investors, compute partners, research collaborators — are inputs to that vendor's regulatory risk profile, not just to its financial profile. When SK Telecom's relationship with Anthropic became the trigger for a government directive, it affected every Anthropic customer globally, not just SK Telecom.

The G7 working lunch is the political context within which the resolution of the Fable 5 situation is most likely to occur. When the leaders of seven major economies convene specifically to discuss AI deployment with the CEOs of the companies affected by export control disputes, the diplomatic signals are clear. President Trump's "going fine" characterisation of negotiations from the G7 setting suggests the resolution is proceeding within a diplomatic framework that extends beyond the bilateral Anthropic-Commerce Department relationship. The presence of Amodei and Altman at the same G7 table — at the same time that both companies are in IPO processes dependent on investor confidence — provides additional incentive for all parties to find a resolution that restores access before the financial market windows that matter most to both companies.

The G7 AI governance dimension goes substantially beyond the Fable 5 situation. The working lunch produced what French diplomatic readouts described as consensus on three principles: that AI deployment should be accelerated for economic competitiveness, that safety and governance frameworks should not prevent responsible deployment, and that international AI governance coordination should be prioritised over unilateral national restrictions where possible. The third principle is the most consequential for the Fable 5 resolution — and for every future enterprise AI vendor risk scenario. A G7 consensus that international coordination should be prioritised over unilateral restrictions creates political pressure against exactly the kind of single-country export control directive that produced the June 12 shutdown.

For enterprise technology leaders watching the AI governance trajectory, the G7 working lunch is the moment that confirms AI model regulation has entered the diplomatic tier. The export control directives, the state legislation, the EU AI Act and the Great American AI Act we have been covering throughout this year are now accompanied by a multilateral diplomatic framework that the world's most powerful economies are building in parallel. The enterprises that track that diplomatic framework — and its implications for the availability, pricing and governance requirements of frontier AI models — will make more accurate long-term AI platform decisions than those who treat regulation as a purely domestic compliance concern.

At Legacies Techno, the G7 working lunch and the SK Telecom trigger have two specific implications for our client engagements. Our AI-Powered Platforms practice is updating its vendor risk assessment templates to include strategic partner due diligence as a standard input — specifically, evaluating which governments have raised concerns about which of a vendor's investors, research partners or compute supply relationships. The SK Telecom trigger demonstrates that a vendor's partner relationships are a material input to the regulatory risk profile of that vendor's most advanced models.

Our Enterprise Software Development practice is specifically noting the Amazon vulnerability disclosure dimension of the Fable 5 trigger. Amazon researchers identified potential vulnerabilities in Fable 5 and reported them to the White House, contributing to the chain of events that produced the export directive. Amazon is also AWS, which hosts Bedrock and which Anthropic has made one of its primary distribution partners. The fact that Amazon's researchers contributed to the government action against Anthropic's model while AWS hosts Anthropic's commercial API access is the kind of complex multi-party relationship dynamic that enterprise vendor risk assessment frameworks were not designed for and must now accommodate.

The week that began with the SpaceX SPCX IPO ended with a G7 working lunch on AI governance. Enterprise AI has arrived at the intersection of capital markets, geopolitics and technology regulation simultaneously. The organisations that understand all three dimensions — not just the technology — will navigate what comes next from a position of clarity rather than surprise.

Key Highlights

Why This Matters