Andrew Ng and fellow AI leaders are pushing back against the idea of letting a handful of companies control who gets to use artificial intelligence. At the Ai4 conference in Las Vegas, Ng, Nobel laureate Geoffrey Hinton, and World Labs CEO Fei-Fei Li debated how to balance innovation, safety, and fair access to AI technology.

Why Does Andrew Ng Fear AI Gatekeepers?

Ng expressed concern that if a small number of companies monopolize AI technology, it would severely limit how people around the world can use these tools. He emphasized that competition among different AI providers and models is crucial to prevent any single player from dominating the market.

"I don't want there to be gatekeepers. That limits all of our access to AI," said Andrew Ng.

Andrew Ng, Founder of Coursera

Ng also highlighted a geopolitical dimension to this concern. He warned that inexpensive open-source AI models from China could become widespread across Asia, Africa, and other developing countries. In his view, this matters because it could influence how billions of people receive information about democracy, freedom, and human rights. He called for supporting U.S. competitiveness in open-source AI software development to ensure diverse perspectives shape global AI systems.

How Should the AI Industry Balance Openness and Safety?

The three panelists disagreed on some specifics but agreed that the answer is not simply "all open" or "all closed." Geoffrey Hinton distinguished between two different concepts that are often confused: open-source software and open-weight models. Open-source code allows programmers to examine and modify the underlying instructions, which helps identify bugs and vulnerabilities. Open-weight models, by contrast, release the trained parameters of an AI system, which Hinton argued could make it easier for bad actors to adapt powerful models for malicious purposes like cyberattacks.

However, Hinton acknowledged that open-weight models have already become an established part of the industry. The financial barrier to training large foundation models has effectively disappeared, making it harder to control access through cost alone.

"Open source is wonderful. You show people the code, and lots of people look at the lines of code and say, 'Oh, there's a bug here.' Open weights means you train a large model and then give people the weights. That is very different," explained Geoffrey Hinton.

Geoffrey Hinton, Nobel Laureate

Fei-Fei Li proposed a nuanced approach inspired by how nuclear physics is regulated. Scientific papers are published openly, but uranium is strictly controlled. Similarly, different parts of the AI ecosystem could have different levels of openness.

"I think we should use AI as this kind of infrastructure. We need certain levels of openness in scientific discovery, education, and global partnerships, as well as profitable business models for entrepreneurs. But we will also embrace closed-source systems," stated Fei-Fei Li.

Fei-Fei Li, CEO and Co-founder of World Labs

Ways to Achieve Balanced AI Governance

• Tiered Openness: Allow different components of the AI ecosystem to have varying degrees of openness, with scientific research and education remaining accessible while certain commercial products remain proprietary.
• Competitive Markets: Maintain multiple AI providers and models to prevent any single company from becoming a gatekeeper that controls who can access and use AI technology.
• Regulatory Frameworks: Implement rules that guide AI development toward benefiting society, with particular attention to safety testing and transparency before powerful models are released.
• Creator Compensation: Establish licensing systems that allow authors, artists, and creators to determine whether their work can be used for AI training and to negotiate fair payment for that use.

All three panelists agreed that some form of regulation is necessary to ensure AI development benefits humanity rather than concentrating power in the hands of a few large corporations. Hinton stressed that regulation should function as a steering system, not brakes, directing AI development toward positive outcomes rather than halting progress entirely.

The debate reflects a broader tension in the AI industry. Ng noted that some large technology companies may exaggerate fears about job displacement to boost their market position, while simultaneously claiming that AI will create enough new jobs to offset losses. This contradiction, he suggested, leaves the public confused about what AI will actually mean for employment and society.

The stakes are particularly high because AI's capabilities are advancing rapidly. Hinton noted that AI's ability to complete tasks effectively doubles every seven months, and he expects massive disruption of white-collar work by the late 2020s and early 2030s. Against this backdrop, decisions about whether AI remains open or becomes controlled by a few gatekeepers will shape how billions of people interact with the technology for decades to come.