It’s Time to Bring Back Congress’s Nerd Squad

Congress once had an office for making sense of new technologies. It may be time to revive it in the age of AI.

In the autumn of 2001, a new national security threat arrived by mail. Envelopes carrying a fine, nearly invisible powder had begun traveling through the postal system. One letter reached a tabloid’s newsroom in Boca Raton, Florida, killing a photo editor within a few weeks. Other letters arrived at NBC News and the New York Post. In mid-October, a letter addressed to Senate Majority Leader Tom Daschle was opened in the Hart Senate Office Building, releasing anthrax spores and shutting down the building for months. By the time the letters stopped, five people were dead and 17 more infected.

Still reeling from September 11, lawmakers scrambled. How easily could a biological weapon be acquired or made? What might happen if the attack were larger and more sophisticated?

At a congressional hearing that November, lawmakers cited findings from a 1993 report titled “Proliferation of Weapons of Mass Destruction: Assessing the Risks.” Written in plain language, the report described various weapons’ effects, the technical aspects of monitoring and controlling production, and the trade-offs among nonproliferation tools. It happened to specifically warn that under the right weather conditions, 100 kilograms of anthrax released in the nation’s capital could kill up to three million people—a casualty outcome that, “pound for pound of weapon,” could exceed a nuclear bomb’s. According to one journalist, “in the rush to sort out fact from fiction, rumor from real scientific fact, Congress was using reliable information from a group that was charged with studying the real ramifications of bioterrorism.”

That group, the Office of Technology Assessment, or OTA, had in fact been defunct for years. Created in 1972, OTA had been a small but proud agency that worked for Congress alone, helping lawmakers navigate science and technology issues with nonpartisan expertise. When advances in genetics raised the prospect that a blood test might reveal a disease-causing mutation, it was OTA that examined what that knowledge might mean for privacy, insurance, and discrimination. Over the years, the agency took on acid rain, mammograms, telecommunications, drug abuse, missile defense, intellectual property, pest control, electronic surveillance, agricultural technologies, and the disposal of nuclear waste, among hundreds of other issues. Wherever Congress might be caught flat-footed as the world changed, OTA could be found nearby.

Congress appreciated the help. A 1990 survey of congressional staff reported that zero percent of respondents found OTA “not useful.” (Half of them said “very.”) A 1993 study by the American Enterprise Institute and the Brookings Institution concluded that “the agency is considered highly credible by members of both parties.” Nonetheless, in a mad rush to slash federal spending in the mid-’90s, Congress defunded it. “We are cutting off one of the most important arms of Congress when we cut off unbiased knowledge about science and technology,” lamented Republican Representative Amo Houghton. A few years later, as a biological weapon made its way through the mail, Congress reached for the work of a team it had let fall by the wayside.

Right about now would be a good time to bring OTA back. In recent months, a parade of senior researchers and executives from the country’s top technology companies has issued alarms about artificial intelligence, using terms like “human extinction.” Jacob Coxon, a former researcher at Anthropic, wrote that companies like OpenAI and his own former employer are “gambling with our lives.” The prospect of advanced artificial intelligence escaping human control is not a new worry, but the very public warnings by industry insiders have very suddenly shaken Washington.

Company leaders have their preferred approaches. OpenAI wants mandatory national safety requirements and industry-led standards. Anthropic has called for independent safety evaluators inside companies and a slowdown in development. Nvidia argues that slowing down is unnecessary and that speed and safety are a “false choice.” The executive branch, meanwhile, has its own technical experts, while lobbyists, trade associations, academics, labor groups, and AI safety organizations press their own theories of problems and solutions. Congress is hearing from no shortage of people with something to say about the way forward.

The issue facing Congress, at least initially, is not one of selecting a particular approach among a clear set of options. It is to determine which claims being made are true, what remains unknown, and what each proposed........

© New Republic