Should AI Be “Born Secret”?

Kevin O'Toole
AI: Purpose Driven Policy
4 min readAug 20, 2024

“When the time comes to hang the last capitalist, he will sell us the rope.”

-Vladimir Ilych Lenin, father of Communism

As the Cold War began, Congress took action to address the dual threats of nuclear weapons and Soviet expansionism. The resulting 1946 Atomic Energy Act laid broad foundations that the government built upon over several decades. The act included a sweeping provision that nuclear research, designs and technology were “born secret.” That is, anything related to nuclear technology was immediately classified as a national security secret. It didn’t need to be part of a specific classified program. By its very nature the government considered (and still considers) nuclear technology an issue of national security.

Fast forward 80 years and the country finds itself wrestling with the dual challenges of the AI revolution and the rise of an authoritarian China that has demonstrated both intent and prowess in applying AI for a variety of nefarious ends. Operating within the confines of existing law and executive powers, both the Trump and Biden administrations have worked to slow China’s access to advanced AI technologies. Most notably, the government has focused on restricting access to powerful AI processors from Nvidia and other companies.

The Chinese passion for acquiring these advanced chips is itself evidence of their strategic importance. For 50 years the US military has commanded a lead in the information domain. This lead in designing, building, operating and networking everything from sensors to weapon systems has helped secure Western democracies for generations. Losing this lead — or even allowing China to remain close behind — has wide ranging and dangerous implications.

Against this backdrop, an excellent and disconcerting article from the New York Times merits particular attention.

The article’s authors detail the lengths to which China is going to obtain advanced AI technology. The quiet complacency of American tech companies — which have trillions of dollars of profits on the line — is disconcerting at best and dangerously self-serving at worst.

The New York Times article reports that American companies, while appearing to comply with export restrictions, often skate as close to the legal line as possible. After China used Nvidia’s A100 chips to accelerate their hypersonic weapons efforts, the US government placed restrictions on the sale of those chips. Nvidia quickly made modest changes to the chips and began selling the new version under the label A800. Meanwhile, China formed a company with Nvidia, MSFT and Intel to ship AI to China. In another instance the executives of companies placed on export restriction lists simply formed a new company and began buying the same advanced AI capabilities from the key tech players.

Can anyone imagine Cold War executives at Boeing and Lockheed having formed a JV with the Mikoyan company (maker of MiG jets) to help transfer nuclear technology to the Soviet Union? Or the government’s reaction if General Dynamics shared nuclear secrets with a new Chinese company whose executives had just left a different company on the banned entities list?

China is investing heavily into its military. The timing and full scope of their intentions remain unclear but the investment trajectory does not. They are surging development of naval and stealth aircraft capabilities — often based on stolen US designs — with the explicit goal of being able to stare down the US Navy. The Chinese Navy is well on its way to being a true “blue water” force that rivals the US in a way the Soviets never dreamt of. Within a few years Chinese Naval formations will undoubtedly routinely sail off of US coasts. They have brazenly built full scale target outlines of US aircraft carriers to facilitate development of their DF-21 “carrier killer” ballistic missiles. The nation must take these risks seriously. China’s access to advanced AI capabilities may well determine whether western democracies are able to maintain a military advantage that helps avert a destructive regional or global conflict.

Every so often US citizens are reminded of the country’s deeper governing structures. Close elections remind the country that it is a Republic of States, not a direct national democracy. Bills that stall in the Senate highlight a bi-cameral legislature that balances the powers of large and small states within this sprawling republic. Supreme Court rulings limit executive freedom and highlight the separation of powers.

A reminder about the nature of national security and great power competition appears to be in the offing. Through the course of the Cold War several people were imprisoned for disclosing nuclear secrets to the Soviets. Julius and Ethel Rosenberg were even executed — the only people ever to be so under a WW I treason law.

To protect its national security interests, the country may have to decide that, like nuclear technology, AI research must be born secret. If so, words like “national secrets” and “treason” will enter the AI development conversation.

(Image Credit: Created using OpenArt)

--

--

Kevin O'Toole
AI: Purpose Driven Policy

I write about the need to develop national purpose and governance related to Artificial Intelligence.