Loading the Audio Player…
The co-founder and CEO of Google DeepMind called on the United States to establish an artificial intelligence watchdog organization. In a Substack post Tuesday morning, Demis Hassabis laid out the framework for a new regulatory body. Such an organization would hit the brakes if most advanced AI models become too dangerous, said Hassabis, who is also a Nobel laureate. Regulation is urgently needed, he said, claiming that AI systems with the cognitive capabilities of the human brain are only a few short years away. If the United States took the first steps of implementing safeguards, other countries would follow suit, said Hassabis, a British citizen who has advised the U.K. government.
Hassabis told Axios that he had spent months building support for his proposal by briefing the Trump administration, other AI labs, and European officials. He hopes to see the organization established in the U.S. before the end of the year, he added.
How would the AI watchdog organization function? An organization could be set up to resemble existing regulators like Wall Street’s Financial Industry Regulatory Authority, Hassabis said. It would be manned by the leading independent experts and funded by the AI industry, in “a federally overseen public-private partnership,” he said. Financial Industry Regulatory Authority describes itself as a self-regulatory organization that polices its member organizations while under oversight from the U.S. Securities and Exchange Commission.
The his proposed regulatory organization would be responsible for developing assessment protocols and working with federal agencies to evaluate advanced AI models, Hassabis said. Frontier AI technology would be tested for dangerous cyber, biological, and deception capabilities, he said. The organization could also implement an industry-wide slowdown if needed. What the world does now to regulate AI will determine how a new phase of civilization unfolds, Hassabis said.