• A crackdown that set the tone
• G7 debates “trusted partners” — and US leadership
• China courts the Global South with free models
• Tech industry pushes “trusted” platforms

G7 Leaders Discuss Global AI Governance and Access World leaders left this week’s G7 summit united in wanting cutting‑edge American AI — but increasingly worried that Washington alone controls the off‑switch.

§A crackdown that set the tone

Days before leaders met in Évian-les-Bains, the US Commerce Department ordered Anthropic to halt exports of its newest Mythos 5 and Fable 5 models on national security grounds, abruptly cutting off all foreign users. The move turned a longstanding fear into a concrete risk: any country or company building on US AI infrastructure could see access “revoked overnight, for reasons they may never be told.”

At the summit on Wednesday, France’s Emmanuel Macron and India’s Narendra Modi warned that if the US “from one day to the next can turn off the switch,” it could damage both their economies and the very AI firms they rely on, according to reporting from the meeting.

§G7 debates “trusted partners” — and US leadership

In closed‑door talks, G7 leaders explored a “trusted partner” scheme to guarantee continued access to American models while accepting tighter US export controls. Anthropic CEO Dario Amodei urged governments to “resist the temptation to splinter” AI governance and maintain international co‑operation.

Over lunch, Amodei and Google DeepMind’s Demis Hassabis went further, proposing a US‑led coalition to set AI rules and standards and to coordinate on access to frontier models and trade in critical chips that would exclude China. Canada’s Prime Minister Mark Carney signalled support for the US taking the lead, though the gathering ended without binding commitments.

AI researchers watching from the sidelines warned that this path risks “walled AI gardens where we beg for access from a few East India companies,” with open source potentially “banned on national security grounds,” a scenario amplified in a widely shared post boosted by Meta AI chief scientist Yann LeCun.

§China courts the Global South with free models

As the G7 debated controlled access, Beijing unveiled a rival vision. China announced it was “accelerating the establishment of a global AI cooperation organization” and invited all countries to join. Its approach centres on cheap or free open‑weight models like DeepSeek and Qwen that can be fully downloaded, in contrast to US subscription‑only systems under growing export curbs.

Chinese officials attacked “closed, exclusive and monopolistic approaches to tech development,” explicitly contrasting their offer to the G7’s “trusted partner” idea and targeting developing nations priced out of enterprise AI subscriptions. They pointed to initiatives such as “AI Capacity Building for All” and cooperation channels via BRICS and the Shanghai Cooperation Organisation to embed AI support for the Global South.

§Tech industry pushes “trusted” platforms

Some industry leaders framed the G7 talks more optimistically, emphasising joint work on “trusted AI platforms.” Salesforce’s Marc Benioff, referencing discussions with Hassabis, European Commission President Ursula von der Leyen and Mistral AI’s Arthur Mensch at the summit, highlighted efforts to “innovate, grow, & develop infrastructure with the fundamental basis of trusted AI platforms.”

For now, however, the world is edging toward two structural blocs: a US‑centred system of tightly governed, paywalled AI and a Chinese‑backed ecosystem of lower‑cost, more open models. Between them, the Global South and mid‑sized democracies are pressing to avoid becoming mere tenants in someone else’s digital empire.

Continue reading https://foxvector.com/stories/019ed886-1356-3a17-703c-030e1783ccc7