Three prominent AI researchers addressed AI safety concerns at the AI4 conference in Las Vegas, emphasizing the role of open models. As major labs seek ways to ensure safety, open-weight models have raised questions about control and misuse. Ng said he opposed gatekeeping in AI and advocated for openness to allow broader access, noting that concentrated control by a few firms could limit innovation and create competitive disadvantages. He suggested multiple providers compete rather than a small number dominating the field. Hinton distinguished between open-source software, where code is publicly viewable, and open-weight models, where trained parameters are released. He argued that releasing model weights makes it easier for actors to adapt large models for harmful purposes such as cyber attacks, and said competition to develop cheaper models could shift market dynamics, especially if nations such as China achieve more cost-efficient AI production. Hinton acknowledged that open-weight models are now permanent and that the cost barrier for accessing large models has largely disappeared. He expressed optimism that AI advancement would improve productivity, education, and healthcare, rejecting characterizations of safety concerns as fear-mongering. He called for regulation to guide AI development, stating that decisions should not rest solely with technology executives. Li emphasized the need for nuanced openness, arguing that a binary view of total openness versus total closure is misleading. She cited scientific practices, such as publishing research while regulating hazardous materials, and pointed to the Human Genome Project as an example of how shared knowledge can support public benefit and commercial innovation. All participants agreed that some form of regulation would be necessary to ensure AI serves public interests and that exclusive control by a few companies could hinder broader societal benefits.