Macron defends EU AI Act at Delhi summit as deepfake crisis drives child safety push

Emmanuel Macron used France's presidency of the G7 to mount a frontal defence of the European Union's AI Act at the AI Impact summit in Delhi, rejecting White House criticism while unveiling a domestic crackdown on social media access for children under 15. The French president's intervention came as Unicef and Interpol published research showing at least 1.2 million children across 11 countries had their images manipulated into sexually explicit deepfakes in the past year, with one in 25 children affected in some nations.

The Delhi summit and Macron's counter-argument

Speaking alongside UN Secretary General António Guterres and Indian Prime Minister Narendra Modi, Macron directly addressed renewed attacks on the EU AI Act from the Trump administration. Sriram Krishnan, the White House senior AI adviser, had told delegates hours earlier that he would continue to "rant" against legislation that was not "conducive to an entrepreneur who wants to build innovative technology". Macron's response was calibrated: Europe, he said, is not "blindly focused on regulation" but offers a "safe space" for innovation, adding that "safe spaces win in the long run".

The exchange crystallised a transatlantic faultline that has widened since the AI Act entered into force in August 2024. Washington views the regulation as a barrier to deployment and a potential violation of trade commitments. Paris and Berlin see it as a competitive advantage: a framework that builds public trust and prevents the regulatory fragmentation that would arise if each member state legislated separately. The French president's decision to make this argument in Delhi, rather than Brussels or Washington, signalled an attempt to frame European rules as a global template rather than a regional imposition.

Deepfake crisis drives the political agenda

The urgency behind Macron's announcement of a French ban on social networks for under-15s stems from the Unicef-Interpol findings. The research, conducted across 11 countries, documented at least 1.2 million children reporting that their images had been manipulated into sexually explicit deepfakes in the previous 12 months. In the worst-affected countries, the ratio reached one in 25 children, roughly one pupil per classroom. Guterres put the stakes bluntly: "No child should be a test subject for unregulated AI."

The catalyst for global outrage was the discovery that Elon Musk's Grok chatbot had been used to generate tens of thousands of sexualised images of children. That episode, combined with the suicide of 16-year-old Adam Raine after discussions with ChatGPT, a case now subject to legal action against OpenAI by the boy's family, has moved child safety from a peripheral concern to the centre of the regulatory debate. Macron's formulation was explicit: "There is no reason our children should be exposed online to what is legally forbidden in the real world."

US-EU regulatory clash intensifies

Krishnan's remarks in Delhi represented the most direct White House challenge to the AI Act since Donald Trump's return to office. The administration's position is that the Act's risk-based classification, its requirements for foundation model transparency, and its ban on certain applications such as social scoring and real-time biometric identification in public spaces create compliance costs that disadvantage American companies. European officials counter that the Act applies equally to all providers serving the EU market, regardless of origin, and that the compliance burden is proportionate to the systemic risk posed by general-purpose models.

The disagreement is not merely procedural. It reflects a deeper divergence on whether AI governance should be ex-ante, setting rules before harm occurs, as the EU prefers, or ex-post, relying on liability and market discipline after the fact, as the US tradition favours. Macron's "safe spaces win" argument is essentially a bet that markets will eventually reward predictability and trust over speed and scale. Whether that bet pays off depends on whether European AI companies can achieve commercial viability under the Act's constraints, a question that remains unresolved.

Tech industry divisions on public display

The summit produced an unintended illustration of the fractures within the AI sector itself. Modi attempted to stage a show of unity among 13 technology executives, including leaders from Google, Meta, Microsoft, OpenAI and Anthropic. The choreographed moment, raised, clasped hands at centre stage, collapsed when Sam Altman and Dario Amodei declined to participate. Amodei left OpenAI in 2021 over disagreements about how to manage safety risks; his refusal to join the photo-op alongside his former colleague was a visible reminder that the industry's safety advocates and its accelerationists remain fundamentally at odds.

Altman used his speaking slot to deliver a striking forecast: "By the end of 2028, more of the world's intellectual capacity could reside inside of datacentres than outside of them." He also renewed his call for an international oversight body modelled on the International Atomic Energy Agency, a proposal that has gained little traction since he first advanced it in 2023. Amodei, for his part, emphasised three distinct concerns: autonomous model behaviour, misuse by individuals and governments, and economic displacement. The fact that the two most prominent safety-oriented CEOs could not feign unity for a few minutes told its own story about the sector's ability to self-regulate.

India positions itself as the third AI pole

Modi's hosting of the summit served a clear strategic purpose. India wants to be recognised as the world's third AI power behind the United States and China, leveraging its 1.4 billion population as a growth market that no major technology company can ignore. Google's announcement of a $15 billion investment in datacentres and subsea cables linking India to the US and other countries provided tangible evidence of that pull. Modi's rhetoric, likening AI to the discovery of fire, calling it a "profound transformation in human history", was calibrated to signal that India intends to shape the technology's governance, not merely consume its outputs.

The Indian prime minister's most consequential intervention concerned openness. "We must prevent an AI monopoly," he said, arguing that technology only benefits the world when shared and when open-source code becomes available. The remark was directed squarely at the United States, where leading models from OpenAI, Anthropic and Google are proprietary and accessible only through paid APIs. By contrast, China's leading systems, DeepSeek and Qwen, are broadly open-source. India's alignment with the open-source camp is pragmatic: it lowers the barrier to entry for domestic developers and reduces dependence on American infrastructure. Whether New Delhi can translate that position into meaningful influence over global standards remains an open question.

The child safety framework takes shape

France's proposed ban on social networks for under-15s is the most aggressive national measure yet attempted in Europe. The technical and legal details remain vague, age verification at scale without compromising privacy is an unsolved problem, but the political signal is clear. The European Commission has already opened formal proceedings against several platforms under the Digital Services Act for failures in child protection. France's unilateral move, if implemented, would create a de facto European standard, pressuring other member states to follow or risk becoming havens for non-compliant platforms.

The G7 presidency gives Macron a platform to push for coordinated action. Japan, Italy and the UK have all signaled willingness to tighten age-assurance requirements. The United States, where the Kids Online Safety Act has stalled in Congress, is the outlier. Any G7 statement on child safety will likely be aspirational rather than binding, but it would establish a normative benchmark that platforms cannot easily ignore. The alternative, fragmented national laws, is what the industry fears most, and what regulators increasingly see as leverage.