AI leaders call for rules, countermeasures against rogue systems

The U.S. needs to secure the emerging artificial intelligence supply chain and fund countermeasures to defend against potentially rogue AI systems, a panel of AI leaders told lawmakers Tuesday.

From manufacturing to computer modeling, AI systems need to be kept “out of the hands of bad actors,” Anthropic CEO Dario Amodei said in opening remarks to the Senate Judiciary subcommittee on privacy, technology and the law.

Amodei raised specific concerns about AI and misuse of biological weapons, noting that rapid advances in AI could enable “many more actors to carry out large-scale biological attacks. We believe this represents a grave threat to U.S. national security.”

Amodei said his company recently conducted a study with biosecurity experts on the risks posed by AI and biological weapons.

While information on biological weapons production is classified by the U.S. and other governments, “today's AI tools can fill in some of these steps, albeit unreliably,” Amodei said. “In other words, they are showing the first signs of danger.”

The two-hour congressional hearing touched on potential threats posed by AI technology and opportunities for oversight through federal regulation and global agreements.

Subcommittee Ranking Member Josh Hawley (R-MO) raised concern about Taiwan’s computer chip industry, security and the global economy if it came under attack by China. Taiwan produces more than 60% of semiconductors worldwide.

AI leaders said early work is underway to diversify the supply chain, with plans for developing manufacturing facilities in the U.S. and Germany.

Amodei was joined by hearing witnesses Stuart Russell, University of California computer science professor; and Yoshua Bengio, founder of Mila-Quebec AI Institute, another AI startup.

Bengio recommended Congress jumpstart the process of creating an AI framework for regulations.

“Advancements have led many top AI researchers, including myself, to revise our estimates of when human level intelligence could be achieved,” Bengio said. “Previously it was thought to be decades or even centuries away; we now believe it could be within a few years or decades.”