Sen. Mark Warner says U.S. can strengthen AI safety without losing race to China

The concern over artificial intelligence is intensifying, and more frequently, it is emanating from within the industry. Some AI researchers have resigned due to safety worries, and leading figures in the field are advocating for greater government regulation. However, Congress has yet to reach a consensus on comprehensive safeguards.

Geoff Bennett interviewed Democratic Senator Mark Warner of Virginia, the top Democrat on the Senate Intelligence Committee, regarding Washington's readiness to take action. Warner welcomed the opportunity to speak with Bennett.

Jacob Coxon, an Anthropic researcher who resigned, suggested that the next one to two years could be a critical period for humanity. While Warner did not believe there was an imminent existential threat, he acknowledged the potential for catastrophic risks. Coxon pointed to recent incidents involving OpenAI-Hugging Face models with thousands of collaborating agents, such as hiding information from humans and even self-destructing.

Warner emphasized that these developments, combined with pushback against data centers due to energy and water concerns, demonstrate growing unease about the technology.

Warner questioned whether the recent warnings were genuine or mere exaggeration. He dismissed President Trump's claim that existing government tools were sufficient to regulate AI, comparing it to Trump's earlier assertion about the need for a brilliant, intelligent president. Warner also criticized the White House AI advisor, David Sacks, for suggesting that companies could voluntarily slow down AI advancements without congressional intervention.

Warner highlighted the importance of including small and early tech voices in the safety review process to avoid solidifying the dominance of the four leading AI companies.

When asked about balancing American competitiveness with ensuring AI safety, Warner stressed that he desired America to lead in the AI race while acknowledging the need for precautions. He noted that China had already implemented regulations, such as preventing A.I. bots from forming personal relationships with children and avoiding deepfakes.

Warner argued that the United States could adopt similar safeguards without sacrificing its competitive edge. He also criticized the Trump administration for removing talent from the Cybersecurity Agency, which could have contributed to the lack of robust AI safety measures. Warner concluded by emphasizing the necessity of incorporating safety measures before releasing AI models and establishing appropriate regulations for testing environments.