Breaking

Trump, AI CEOs Meet at G7 to Weigh Joint Safety Rules

June 17, 2026 at 13:07 EDT

Policy & Regulation
Foundation Models

On June 17, 2026, at the G7 summit in Evian, France, U.S. President Donald Trump—joined by his secretaries of state, treasury and commerce—convened a working lunch with the CEOs of major AI companies including OpenAI, Google DeepMind and Anthropic to discuss safety standards set by democratic nations.

June 17, 2026 · G7 Summit, Evian-les-Bains

Trump and Top AI CEOs Push a U.S.-Led Global Forum on AI Safety

A "working lunch" on the summit's final day signaled a democracies-first effort to set frontier-AI safety standards — with Washington intent on leading it. "It's going to be the biggest thing ever," Trump said.

Frontier-AI CEOs at the table — OpenAI, DeepMind, Anthropic, Meta, Mistral, Cohere, Salesforce

Cabinet secretaries on the U.S. side — State, Treasury, Commerce

30d

Max voluntary pre-release review window for covered frontier models

Pre-Release Review Window — Cut Before Signing

The voluntary government access period was reduced from the initial proposal to the signed order.

90 days

Initial proposal

→

30 days

Signed order

Executive Order · "Promoting Advanced AI Innovation and Security" · June 2, 2026

Scope

Government-designated covered frontier models

Directives

Create benchmarks; stand up an AI cybersecurity clearinghouse

Export Controls

Already applied to Fable 5, Mythos 5 and others

Supporters say

A unified, light federal framework beats a state-by-state patchwork
A reasonable response to demonstrated cyber threats

Critics warn

Export controls and pre-access could squeeze frontier-model supply to allies
Excessive regulation risks hampering innovation

Still open

No joint-statement text confirmed, and no new model names or release timing announced. How far the U.S.-led international forum concept is actually realized remains the key point ahead.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.