I’m about to graduate from law school. Will AI steal my job?

In a startling 2021 interview, OpenAI CEO Sam Altman predicted that in the near future, artificial intelligence (AI) technologies, such as his company’s ChatGPT, would replace many human workers, generate profits on an unimaginable scale and elevate company founders like him to become the world’s first trillionaires. Altman suggested that tech companies like OpenAI could become more powerful than the U.S. government. He implied that the companies controlling AI technologies may ultimately need their own constitutional systems to govern the utilization of their technologies. 

Altman doesn’t seem to recognize that a form of governance already exists that can determine how AI is deployed and ensure that it does not harm the American public: The Constitution and laws of the United States. 

Congress should act quickly to establish a legal framework putting guardrails on commercial applications of AI. Indeed, Altman’s job-automation vision may be closer than most of us realize. The latest version of ChatGPT recently passed the bar exam. ChatGPT also passed the licensing exam for medical doctors and Google’s coding test for entry-level software engineers.

Many professional workers have begun to ask: Is AI going to steal my job? And this isn’t a problem facing just high-income workers. Everyone from call center operators to talk therapists faces the risk of displacement as the commercial deployment of ChatGPT and its ilk begins.

As a student in my final semester of law school, I can see the writing on the wall. Much of the work I have performed consists of reviewing and drafting legal documents; it is clear to me that much of the work I do will be automated in the next few years.

Perhaps the automation of work, lawyerly and otherwise, will be liberating. As Bertrand Russell once speculated, advances in technology may afford workers greater leisure time as automated processes perform more of their work. Philosopher Aaron Bastani used the term “Fully Automated Luxury Communism” to describe a similar vision, in which machines do all the work and the state provides for its citizens’ material needs.

But the road to such utopian fantasies is fraught with danger. Rapid job displacement and a disorderly transition to automation threaten political and social chaos. The rise of a discontented and unemployed professional class does not bode well for democracy.

It is true that other nations such as China are developing AI systems comparable to our own. The U.S. should aim to win the AI race before our adversaries do. But the best way to win may be to roll out these technologies in an orderly manner that avoids domestic social upheaval.

To establish some control over the pace of automation, Congress should buy out the patents and copyrights owned by OpenAI and companies like it. While private innovation generally benefits society, this technology is simply too powerful to be left in the hands of profit-driven corporations. Leaving AI development to private companies would be like privatizing the World War II-era Manhattan Project, which developed the first nuclear weapons.

It bears noting that the patents and copyrights tech companies hold are legal fictions rather than facts of nature. Indeed, Congress’s authority to issue intellectual property rights originates in Article I, Section VIII of the Constitution, the so-called “Patent Clause.” Congress and the tech companies are free to buy and sell their intellectual property (IP) rights. If the companies were to refuse sale of their patents and copyrights to the government, then in theory Congress could assume ownership of the IP rights unilaterally. Under the Fifth Amendment “takings” clause, Congress would likely need to provide the inventors with “just compensation” — the fair value the inventors would have earned if they sold the patents and copyrights voluntarily. 

Having bought the IP rights behind AI, the U.S. government could embark on its own research effort to develop AI technologies ahead of other nations. Congress could also maintain some control over how the technology is used, including whether it is deployed in a way that eliminates jobs.

Moreover, to curb the threat of mass unemployment wrought by AI, Congress should consider enacting a new set of employment laws that prohibit companies from using AI to replace human workers. Although the precise scope of such laws remains to be fleshed out, Congress should look for guidance from laws already on the books that prohibit non-citizens from holding jobs in the U.S. without proper work authorization.

Rightly or wrongly, Congress already restricts the right of foreign citizens to work here, with the stated goal of protecting American jobs. It should do the same with respect to AI.

The text of the Constitution and leading Supreme Court precedent furnish Congress with wide latitude to regulate or prohibit commercial uses of AI that threaten American employment. Indeed, the Constitution’s Interstate Commerce Clause furnishes Congress with nearly unlimited power to regulate economic activity throughout the land.

Crucially, the First Amendment right to free speech presents no bar to laws setting limits on commercial AI deployment. As the Supreme Court stated in Central Hudson Gas and Electric Corporation v. Public Service Commission of New York, speech in the commercial sphere receives a far lower level of protection than core areas of expression such as art and political discourse.

Indeed, Central Hudson establishes that the government may regulate commercial speech if it can show both that it possesses an important interest in regulating the speech and that the regulation is narrowly tailored to that interest.  

The government certainly could show that it possesses a substantial interest in protecting American jobs from automation. Moreover, by drafting the law narrowly to protect only those professions most at risk of automation, the government could show that the scope of such law is not more extensive than necessary to serve its interest in protecting employment. With the Hudson test satisfied, a new AI labor law would not run afoul of the First Amendment.

In sum, the automation of labor is not inevitable. Empowered by its expansive authority under our constitutional scheme, Congress should act now to prevent haphazard rollout of AI technology that causes more harm than good to American society.

Alexander Loznak is a third-year law student at the University of Oregon. Follow him on Twitter @AlexanderLoznak.