Three AI pioneers advocate open approach as safety debate intensifies

Geoffrey Hinton, Fei-Fei Li and Andrew Ng spoke at the Ai4 conference in Las Vegas in early August about regulation, open-source access to AI and the position of the United States as China continues to advance. The three are widely regarded as authoritative voices in the AI field, but their views on how open AI development should be do not fully align.

The debate took place during Ai4 2026, held from 4 to 6 August at The Venetian in Las Vegas. The conference, founded in 2018, was expected to attract around 12,000 visitors this year, along with more than 1,000 speakers and over 400 exhibitors, making it one of the larger commercial AI events in the United States.

Hinton: safety concerns, but no call for closed systems

Nobel Prize laureate Geoffrey Hinton, who has shaped the AI field for decades and is now known for his outspoken warnings about the risks of advanced AI, also took a nuanced position at Ai4. Hinton acknowledges the dangers but did not argue against openness as such. His stance reflects a tension that is widespread in the industry: how do you weigh the risks of proliferation against the benefits of transparency and shared knowledge development?

Hinton won the 2024 Nobel Prize in Physics for his contributions to the development of artificial neural networks. Since then, he has repeatedly warned of scenarios in which AI systems move beyond human control. That makes his presence on a panel advocating openness notable, but also telling: even critics of unconstrained AI development do not see fully closed systems as an obvious answer.

Fei-Fei Li and Andrew Ng on accessibility and competition

Fei-Fei Li, CEO of World Labs and co-creator of the ImageNet database that underpins the current wave of image recognition, stressed the importance of broad access to AI technology. Li has spoken out on multiple occasions about the societal dimension of AI and the need to avoid concentrating knowledge in the hands of a small number of large companies.

Andrew Ng, co-founder of the online education platform Coursera and former head of AI labs at Google and Baidu, echoed that view. Ng has for years been one of the most vocal advocates of open-source AI. He argues that closed models stifle innovation and restrict access for smaller players, including startups and researchers in emerging economies.

A recurring theme in the discussion was the competitive position of the US relative to China. The question of whether American companies and governments should embrace open-source development or instead restrict it to preserve a technological lead divides the field. Ng and Li appear to reason that openness yields greater long-term returns than protection through closed systems.

The regulatory landscape as backdrop

The debate at Ai4 takes place against a backdrop of growing pressure on legislators in the US and Europe to regulate AI. In the European Union, the AI Act is now in force, with obligations that depend in part on the risk level of an application. Open-source models are subject to different rules under the Act, which adds a further legal dimension to the European discussion on openness.

In the US, comparable federal legislation is still absent. The debate over open-source AI there is partly entangled with geopolitical considerations: some policymakers fear that freely available models could also be used by malicious actors or foreign governments. Others, including Ng, point out that the spread of knowledge is difficult to stop and that attempts to do so primarily disadvantage American players.

Implications for the European and Dutch ecosystem

For European and Dutch parties, the outcome of this debate is anything but abstract. Startups and research groups in the Netherlands rely heavily on open-source models, from Meta's Llama series to smaller specialised models shared via platforms such as Hugging Face. Stricter export restrictions or a shift towards closed systems in the US could affect access to those models.

At the same time, European policymakers are grappling with trade-offs similar to those facing their American counterparts. The AI Act provides a framework, but the practical consequences for open-source development have not yet fully crystallised. The fact that three of the most cited figures in the field are publicly speaking out in favour of openness lends weight to that side of the debate, including on this side of the Atlantic.