Project summary

Europe should have greater access to frontier AI models to evaluate their safety. There is a lot of leverage the EU can use, some of which we have detailed in our country governance series. Project AWARE has been researching the ability of governments to obtain AI assurance from developers using various levers of power. We have created a methodology for assessing what levers a state may have and how to utilize them.

European counterparts have expressed interest in commissioned country studies and exploring whether our approach could work at the EU level.

We would like to do six months of research to find out how Europe can better utilise its combined leverage to gain access to frontier AI models.

This would consist of two Track 1.5 conversations in Brussels and result in a framework that could allow European states to act together on this issue.

What are this project's goals? How will you achieve them?

How can European states obtain more access to frontier AI models by combining their efforts? And what tools could they use to achieve this?

1. Expand our research on Country Governance and Leverage to include the EU as a whole. We can then look at how Europe as a whole can leverage this. This could be through legislation, trade, public contracts etc. We will assess how this can augment what the EU AI Office can currently do under Article 92.

2. Come up with ways to combine this leverage. Whether that be by having all states ask for similar terms in contracts or come to an agreement to allow independent evaluations. Perhaps they can even negotiate as a bloc and allow certain levels of access to models. We can build on our methodology for assessing leverage and how to utilise it.

3. Host two Track 1.5 conversations in Brussels. The first would be with key policymakers from around Europe as well as diplomats and technical experts. We can look at what levers are available and any roadblocks. After more research, the second conversation will look at these methods and how they could be used, possibly conducting a mock negotiation.

4. Publish the European Frontier AI Access and Assurance Framework, laying out what levers we believe can be used and how they can be used, as well as when combining efforts may be beneficial.

How will you know if it worked?

We would consider this successful if we can find levers that can be used, rather than judging it by the number of meetings we host.

Expected outcomes would be:

• Two meetings in Brussels with approximately 15 to 25 people each.

• Speaking with at least 15 different parties that may be relevant.

• At least three methods to obtain collective access.

• A public research paper containing this information.

We will seek outside legal review and further discussions with interested European counterparts.

Best case scenario, we can show European counterparts a lever they can use and they decide to take it further.

How will this funding be used?

We are seeking $120,000 for six months, with a minimum funding threshold of $70,000.

The proposed full budget is:

• Research personnel and institutional analysis: $37,000

• External legal and technical expertise: $16,000

• Two Brussels convenings: $18,000

• Participant travel and accommodation: $19,000

• Programme coordination and engagement: $12,000

• Publications and dissemination: $6,000

• Administration and financial management: $6,000

• Contingency: $6,000

At $70,000, we would retain the core research, both Brussels dialogues and final publication, while reducing staffing, travel and external consultancy costs.

Full funding would allow broader participation, more extensive specialist input and deeper assessment of the proposed mechanisms.

Who is on your team? What's your track record on similar projects?

Abdullah X will lead this project. He is the founder and president of Project AWARE, a think tank dedicated to institutional AI safety.

Our previous work has examined country-level leverage and the legal basis for using it. This project applies that research to the EU.

We have worked with governments on similar issues. We hosted an AI Sovereignty Roundtable with the Observer Research Foundation (ORF) in August 2026. We have also briefed 13 states through our engagement programme and created seven bilateral follow-ups.

We have been speaking to European counterparts and they have shown interest in us doing work for their countries. Some have also expressed interest in us looking into how Europe can combine its efforts to get more access to frontier AI models.

We can utilize AWARE's current network of researchers and add more Europeans to the conversation.

What are the most likely causes and outcomes if this project fails?

We do not expect to struggle with initial institutional interest. European counterparts have responded positively to the idea.

What I am unsure about is whether there will be enough levers that can be used collectively.

Different countries may not be able to use some levers or may want to use different ones. Companies may not want to give more access to these models for evaluation. Some countries may also not see it in their best interest to combine efforts with others.

That's something we hope to research through our meetings and further discussions.

Worst case scenario, we can't find a way for European countries to gain leverage over these companies together. We would still share what we found and why we couldn't combine these efforts.

How much money have you raised in the last 12 months, and from where?

Project AWARE has operated with limited institutional funding, relying substantially on voluntary research contributions and in-kind support.

Despite these constraints, we have maintained an active international research and diplomatic engagement programme.

We are now seeking dedicated funding to consolidate our existing European work into a structured research programme. No funding for this specific project has yet been secured.