Expert SpeakRaisina Debates
Published on Jun 29, 2026
US restrictions on Anthropic’s Mythos-class models provide the geopolitical impetus that EU-India relations long lacked to advance technology cooperation and build interoperable AI ecosystems
The broader backdrop to this month’s discussions at the G7 in Evian and VivaTech in Paris — where India was the “official AI country partner” — reveals a fairly profound shift in global technology governance. While the Strait of Hormuz crisis and the Russian invasion of Ukraine took top billing at the G7, equally significant was the United States (US) government’s recent decision to limit Anthropic’s Mythos-class models to foreigners. In response to the Trump administration’s diktat, Anthropic pulled Fable 5 and Mythos because there was no practical way to limit access to these models to non-US citizens alone. Dario Amodei, Anthropic’s CEO, was in Evian and urged world leaders not to splinter access to cutting-edge models — a request that resonated with some in the Artificial Intelligence (AI) industry.
Regardless of what Anthropic says or does now, the damage has arguably been done. Industry titans like Satya Nadella have bemoaned that AI as a field is being advanced singularly by advancements in LLMs. It is good to see better capabilities emerge from these models, but as Nadella pointed out on X, “a frontier without an ecosystem is unstable”. As these companies declare ever-advancing capabilities, they have also attracted geopolitical attention. In cases like Mythos and Manus AI — for instance, in April 2026, China kiboshed the sale of Singapore-domiciled Chinese startup Butterfly Effect, which had developed ManusAI, to Meta — Great Powers have stepped in decisively to regulate LLM companies precisely because they equate Large Language Models (LLMs) with AI. While LLM companies may convince their parent states to loosen restrictions on global deployment, the perception that these models can be turned off at the whim of the political executive will now be difficult to overcome.
While the Strait of Hormuz crisis and the Russian invasion of Ukraine took top billing at the G7, equally significant was the United States (US) government’s recent decision to limit Anthropic’s Mythos-class models to foreigners
What does this mean for the European Union (EU), particularly in relation to its approach to India on technology cooperation? For France and the EU, there are broadly two messages here
The first is that Europe is no longer a norm-maker on export controls — it is a norm-taker on AI, certainly. The shoots of a go-it-alone approach were already visible during the Biden administration, whose “AI Diffusion” agenda carved up the world into tiers. Although Tier 1 states — which included most Western European states and NATO allies — would have faced few restrictions under the Biden plan, Europe was not an interlocutor in that export control regime; it was a ‘subject’ of it. This is arguably the first time since the post-war multilateral system came into being that Europe has found itself outside the tent on dual-use technology regimes, negotiating its way in. Take any major dual-use regime, obsolete or current, orchestrated by the US during and after the Cold War, and it becomes clear that Western European powers were indispensable partners in their creation and implementation. There are larger structural and political-economic reasons for this beyond geopolitical kinship alone.
Europe as an AI Norm-Taker
For the present moment at least, the EU will be wary of export controls on AI. Trump, and perhaps his successor in the White House, may co-opt NATO partners for technology cooperation on drones, biosecurity, and even AI, but the Rubicon has been crossed with Mythos. No Brussels-based policymaker will forget that the US was willing to restrict Europe’s access to advanced AI models. This episode is likely to shape the EU’s approach to future technology restrictions. Even if the EU is invited to co-create a dual-use regime on AI, it will know that the scope of these restrictions — since the definition of “AI” itself is being reshaped by frontier models, whose precise risks remain unknown — may at any time be turned against itself.
No Brussels-based policymaker will forget that the US was willing to restrict Europe’s access to advanced AI models
This presents an opportunity for India to engage with Europe on technology governance, unmoored from US-centric restrictions or the normative posturing around dual-use technologies that has been a hallmark of EU diplomacy since the fall of the Berlin Wall. Both sides now have a credible reason to build supply chains together in drones, high-end manufacturing (especially precision tools), nanomaterials, and optical equipment. There are enough European companies and, arguably, willing Indian partners to incubate cooperation in these sectors; however, it will take sustained diplomatic efforts in New Delhi and Brussels to build momentum behind such efforts. In the Trumpian world that we all inhabit, the Mythos export control may seem like a here-and-now decision that can be reversed at any time. But it is important to place that decision in the context of widespread concern in European capitals over US technology dependencies — especially on US cloud service providers. European resentment towards US Big Tech and Washington’s exploitation of digital choke points against institutions like the International Criminal Court was palpable at VivaTech.
Beyond Frontier Models
This is before we turn to AI itself. The second message for the EU is that it needs to go beyond frontier models and, to use Nadella’s words, develop an “ecosystem” for AI. If Trump and the US can be accused of going it alone on AI export controls, Europe can also be accused of going it alone on technology regulation. GDPR and the EU AI Act have arguably not done enough to stop the market dominance of US Big Tech. That is because Brussels’ regulatory approach — reflected also in its institutional division of responsibilities — has traditionally seen Big Tech regulation and the geopolitical dependencies induced by US companies as two different concerns. Fines and penalties by EU regulators have neither dented the bottom lines of these companies nor changed the business models of digital platforms. Without interoperable railroads for payments, identity or health, Europe is still legitimising winner-take-all markets, where the dominant platform is very difficult to displace because it uses non-interoperable technical standards and contractual agreements to prevent data and value sharing, whether upstream or downstream.
This attitude towards digital and AI regulation is shifting in Europe, if discussions at VivaTech and Bharat Innovates are any indication. European policymakers seem genuinely committed to building digital public infrastructure (DPI), including AI-native rails. The reason for this change is that DPI in Europe was previously framed as being in tension with market freedoms and privacy principles. The “P” in DPI was wrongly conflated with state ownership or management of data rails, and there was some reluctance to assume the regulatory burden of a common standard or platform for data sharing across countries
This time around, DPI development is motivated by the need to prevent walled gardens from emerging around AI and to reduce geopolitical dependencies. It is easy to see how a frontier model can lock in data providers through exclusive Service Level Agreements and lock out value and innovation across the rest of the market. This is an especially important concern for Europe. For over seventy years, EU regulators and public sector units have collected and maintained high-quality, high-signal-to-noise-ratio data on critical services in energy, education, agriculture, and labour. This data must be accessible to any innovator who wants it. Put differently, data must be decommodified and not captured solely by frontier model companies.
Cooperation is advancing at a fair clip; however, it will need political momentum and support from the leadership in Delhi and Brussels to be incorporated into the EU-India Strategic Agenda for 2030
An ecosystem-driven approach, centred around DPI and common data rails, can promote competition, give EU startups a fair shot at innovation, and even bring the EU back into the technology governance tent. The reason European states were able to develop dual-use regimes with the US is that trans-Atlantic supply chains (especially in biosecurity and applied physics) had become so interwoven that export controls were inconceivable without European companies. Walled gardens developed by Big Tech essentially concentrated those supply chains around a single US-based actor and away from European businesses.
A New Agenda for EU–India cooperation
India faces a similar challenge. India has a long history of building DPI for the domestic market, but it now needs to co-create transnational rails that enable cross-border data-sharing. At the AI summit in New Delhi this February, India and France announced a pilot to facilitate the sharing of health data between ICMR in India and the Health Data Hub (Plateforme des Données de Santé) in France. Using India’s Data Empowerment and Protection Architecture, both sides will share health data with innovators to advance AI-enabled diagnostics and genomic mapping. This is a consequential initiative. It signals to other European states that France is prepared to share data with Indian institutions in a highly sensitive domain such as health. The expectation is that this pilot can establish a model of data governance built around DPI frameworks (such as DEPA, or whatever standard India and Europe develop together) that facilitates the development of AI-enabled specialised services in critical sectors.
On this second front, cooperation is advancing at a fair clip; however, it will need political momentum and support from the leadership in Delhi and Brussels to be incorporated into the EU-India Strategic Agenda for 2030. Discussions at Bharat Innovates and VivaTech have made it amply clear that there is broad geopolitical convergence between the EU and India in their approaches to technology governance today. The shift in the European approach is dramatic and would have been difficult to fathom even a few years ago. India should help transform this moment of convergence into sustained cooperation on AI.
Arun Sukumaris an assistant professor at Ashoka University and a non-resident fellow at ORF Middle East
The author acknowledges using Claude AI to fact-check this article
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
PREVNEXT
Author
Arun Sukumar
Read More +
Related Search Terms
Publications
Evaluating the AI Threat to India’s IT Model
Cyber and Technology | Artificial Intelligence
Jun 29, 2026
