G7 pushes 'trusted partners' scheme to access US AI models

What happened?

On Monday, during the inaugural dinner of the G7 summit in the French town of Evian-les-Bains, representatives from several member countries raised with US Commerce Secretary Howard Lutnick the creation of a 'trusted partners' program to allow allied nations to access the most advanced artificial intelligence models developed on US soil. The conversation, according to The Next Web, took place on the sidelines of the dinner, as is customary at these summits, and reflects the growing concern among allies about falling behind in the AI race. The proposal seeks to establish a regulated channel that guarantees security and reliability in the transfer of AI technology, preventing it from falling into the hands of hostile actors. Although there are no concrete details yet, the initiative is based on the premise that G7 countries — France, Germany, the United Kingdom, Italy, Japan, Canada, and the US itself — need a common framework to share these strategic assets without compromising national security.

Why is this important?

This move is part of the growing geopolitical tension surrounding artificial intelligence. The United States has imposed restrictions on the export of chips and AI models to certain countries, especially China, deeming them a threat to national security. In October 2022, the Biden administration introduced export controls on advanced semiconductors, and in 2023 it expanded these restrictions to frontier AI models. However, its G7 allies need access to these models to maintain their competitiveness in key sectors such as healthcare, defense, and industry. For example, France has invested 2.5 billion euros in its national AI strategy, and Germany has allocated 5 billion euros by 2025, but both lack their own frontier models. The 'trusted partners' scheme could be a mechanism to bypass these restrictions without compromising security, establishing a framework of mutual trust. Moreover, this proposal comes at a time when the European Union is advancing its AI Act, which classifies systems according to risk, and where countries like the United Kingdom have promoted the Bletchley Declaration on AI safety. The G7 initiative thus seeks to align security interests with the need for collaborative innovation.

Consequences and outlook

If realized, this program could transform the global AI landscape. On one hand, it would consolidate US leadership in developing frontier models, while allowing its allies to benefit from them without having to develop their own capabilities from scratch. Companies like OpenAI, Google DeepMind, and Anthropic would see their potential market expand, but they would also face greater compliance and government oversight requirements. On the other hand, the program could generate friction with non-included countries, such as China, India, or Brazil, which might interpret it as an exclusionary technology bloc. This would fuel the debate on international AI governance, similar to what happened with nuclear arms control or the non-proliferation of dual-use technologies. Additionally, the 'trusted partners' scheme would require verification and compliance mechanisms, which could slow down technology transfer. In the business sphere, European and Japanese startups could benefit from access to advanced models, but they would also become more dependent on US infrastructure, raising questions about digital sovereignty. The proposal, though embryonic, signals a shift toward more structured AI governance, where trust and security will be key currencies.

What readers should know

The concept of 'trusted partners' is not new in the technology and defense fields. For example, in the US defense procurement program, there is the ITAR (International Traffic in Arms Regulations) system that regulates the export of military technology. Also in the semiconductor field, the CHIPS for America consortium has established criteria for sharing technology with allies. Applied to AI, this concept represents a qualitatively different step because AI models are not just software; they incorporate data, algorithms, and strategic knowledge. To access them, allied countries will need to meet certain standards of security, transparency, and responsible use, similar to those required by the EU AI Act. Although the proposal is in an embryonic phase, its mere discussion at the G7 indicates that the US administration is willing to explore avenues of controlled collaboration. Readers should closely follow the evolution of this initiative, as it could define global access to artificial intelligence in the coming years. Furthermore, debates are likely to arise over the criteria for being considered a 'trusted partner', oversight of misuse, and the role of technology companies in this new framework. The 2025 G7 summit in Canada could be a milestone for finalizing these agreements.