EU AI Act faces global backlash at New Delhi summit as rivals embrace lighter regulation

When the first global AI safety summit convened at Bletchley Park in late 2023, Europe was the standard-setter. The bloc was finalising the world's first comprehensive artificial intelligence law and its officials helped frame the international conversation on how to govern the technology. Two years later, the mood in New Delhi could hardly have been more different. Across a week of speeches, side events and corridor conversations, the EU's Artificial Intelligence Act was held up not as a model but as a warning, an example of how not to regulate.

A chorus of criticism from Washington and industry

The most direct assault came from the White House. Sriram Krishnan, senior policy adviser on artificial intelligence, used a Wednesday side event to tell European listeners that the atmosphere in the EU needs to change and be more focused on innovation, less focused on governance and less focused on doomerism. He added that the EU law is not really conducive to an entrepreneur who wants to build basic technology. The language was undiplomatic by summit standards and signalled a US administration convinced that Europe has mistaken precaution for leadership.

Industry representatives were no gentler. Amanda Brock, chief executive of Open UK, which advocates for open-source AI, said the EU had shot itself in the foot with the AI Act. She argued that the bloc moved too fast, too early: you can't make regulations around something you don't understand and expect it to be successful. That view, that the legislation was written before the technology's trajectory was clear, was echoed in multiple conversations with startup founders and venture capitalists on the summit's margins.

India charts a different course

The host country offered the clearest alternative. India arrived at the summit having announced a light-touch governance system, a wait-and-see approach that intervenes only on specific, identifiable harms such as deepfakes. The model is emblematic of the path many non-European nations are taking: sector-specific rules applied after harm emerges, rather than a horizontal framework applied in anticipation of it. For a government keen to position India as an AI destination for global capital, the contrast with Brussels was deliberate and visible.

Other delegations from the Global South voiced similar instincts. Officials from Brazil, Indonesia and Kenya described regulatory strategies built on monitoring and targeted intervention, not pre-emptive classification of risk tiers. The EU's insistence on a single, economy-wide rulebook looked, to many attendees, like a solution designed for a European institutional architecture that does not exist elsewhere.

Implementation struggles at home

The criticism from abroad is compounded by friction within the bloc. Since the AI Act passed in 2024, EU authorities have been threading the needle between implementing the flagship law and rolling back safety provisions after complaints from European companies that the requirements were too burdensome. The very fact that revisions are underway before the first compliance deadlines have fully bitten suggests the original text may have overreached. Several national regulators have privately warned that the administrative load on small and medium-sized enterprises could drive AI development to jurisdictions with lighter oversight.

This tension, between the ambition to be the world's AI regulator and the need to attract investment, was visible in the Commission's own summit scheduling. At the 2025 Paris summit, Ursula von der Leyen, president of the European Commission, was a headline speaker. In New Delhi, the Commission's on-stage appearance was relegated to Friday afternoon, when many international attendees had already departed. The symbolism was not lost on delegates: the EU's regulatory moment has passed; the deal-making moment belongs to others.

A defensive European response

European officials in Delhi did not go unheard, but their arguments struggled to land. Henna Virkkunen, the Commission's executive vice-president for tech sovereignty, security and democracy, said in pre-circulated remarks that for Europe, leadership in AI means combining investment, scale and responsible governance. She pointed to the Commission's plans to build AI factories on the continent to ensure sovereign compute power and to a voluntary code of practice for the most advanced models such as OpenAI's GPT and Google's Gemini. Virkkunen declined to respond publicly to Krishnan's remarks.

Emmanuel Macron, one of the few European leaders to attend, attempted a defence in his Thursday keynote. Europe is not blindly focused on regulation, he insisted. But the line drew polite applause at best. French officials later acknowledged privately that the president's speech had been drafted before the full force of the summit's criticism became clear, and that a more substantive rebuttal would be needed in Brussels.

Brando Benifei, the Italian MEP who co-led the Parliament's work on the AI Act, dismissed the US position as quite isolated while admitting there's a different appetite for levels of regulation across the world. The concession, that the EU's regulatory appetite is an outlier, was perhaps the most revealing moment of the European pushback. Even the law's architects now accept that the Brussels effect, the phenomenon by which EU rules become global defaults, may not replicate in artificial intelligence.

Where convergence might still emerge

Not all observers wrote off European influence. Amber Sinha, executive director of the digital rights group EDRi, said the EU may get traction with its voluntary guidelines advising companies on how to file risk assessments for the most powerful models. But I don't see a lot of convergence besides from that, he added. Mark Surman, president of Mozilla, offered a similarly qualified endorsement: the EU's AI Act as well as its GDPR data protection rules are the right types of standards but they're really facing an uphill battle. In the current environment of global digital technologies, it's hard for one region, when things are so polarized, to really set those standards.

That polarization, between a US administration focused on speed and dominance, a Chinese state-led model, and a European model built on rights and risk, was the summit's underlying theme. The EU's bet remains that its standards will eventually be adopted because access to the single market compels compliance. But the New Delhi mood suggested that many countries are willing to forgo that access, or to build parallel ecosystems, rather than import a regulatory framework they see as stifling.

The Brussels effect under pressure

The Brussels effect worked for data protection because the GDPR arrived when digital markets were still forming and the EU was the largest regulatory prize. Artificial intelligence is different. The technology is moving faster than any legislative cycle; the largest models are trained in the United States; the capital is overwhelmingly American; and the geopolitical competition is framed in security terms that make regulatory alignment a strategic choice, not a commercial necessity. India's light-touch model, replicated across much of the Global South, offers a viable alternative for countries that want AI investment without European compliance costs.

European officials know this. The AI factories programme, the voluntary code for frontier models, the Compute Fund, these are attempts to supply the investment and infrastructure side of the equation that the regulatory side alone cannot. But they are playing catch-up. The US Chips and Science Act, the Inflation Reduction Act, and a venture capital ecosystem an order of magnitude larger than Europe's have already set the terms of the race. The AI Act, once touted as the global gold standard, now looks to many in New Delhi like a handicap the EU imposed on itself.