According to Bloomberg, AI pioneer Fei-Fei Li has called for more oversight of AI by independent institutions and the public sector. She believes that safety assessments for increasingly powerful AI systems should not be entirely left to the companies that develop them.

Assessments Should Not Be Left to Developers Alone

As a professor at Stanford University and co-founder of World Labs, Li stated on Tuesday that while AI companies' internal researchers can evaluate the capabilities and safety of individual models, this cannot replace broader standards set by governments, industries, and academic institutions. During an interview with Bloomberg Television, she said: "I still believe it is very important for independent institutions and the public sector such as academia to establish benchmarks. At the same time, the industry and government should also share responsibility."

As AI systems grow stronger, concerns about their potential risks are increasing. Some leaders in Silicon Valley have warned that the rapid development of this technology may eventually go beyond human control. In recent weeks, Dario Amodei, CEO of Anthropic, and Sam Altman, CEO of OpenAI, have both called for slowing down the development of cutting-edge models. The two will discuss AI safety issues with the UN Security Council on Wednesday.

Li said that the existential risks brought by AI ultimately depend on how humans choose to develop, govern, and use the technology. She emphasized that this principle also applies to technologies beyond AI — from electricity and steam engines to modern transportation, society has always needed to balance the benefits of technology with personal safety. She pointed out that humanity "has the power over technology," and "we must maintain control. AI is not about AI itself, but about humanity. It concerns our shared and individual dignity and autonomy, as well as our responsibilities."

Humans Must Maintain Control, Debate Intensifies

Li's remarks add new voices to the growing debate between policymakers and AI industry leaders on how seriously to take warnings about the most extreme risks of this technology. Earlier on Tuesday, U.S. President Trump dismissed concerns about AI's existential threat during his speech at the United Nations General Assembly in New York, calling it "a scam" and emphasizing the need to continue developing the technology at full speed.

Li also pointed out that discussions about AI system behavior have gone beyond so-called "frontier models" that are driving the current boom. Her company, World Labs, is developing "world models" capable of generating and reconstructing 3D environments, which can be used to train and evaluate robots and empower applications in design, healthcare, and other industries. She said that as these systems move into real-world applications, the industry faces the same fundamental question: how to assess increasingly powerful AI systems and determine whether they operate as intended.