Ban Non-Competes, Enable AI Talent
If the FTC's non-compete ban fails, Congress should take up the cause.
Finally free to build. Image Source.
This week, the Federal Trade Commission issued a potentially landmark new rule that if upheld will nullify practically all existing and future non-compete agreements. Already, however, it seems FTC Chair Khans’ ambitious attempt to ‘go big’ on reform will end up in courts. Multiple lawsuits have been filed, and the effort may die in the process. Regardless of jurisdiction fights, the substance of the ruling is profoundly important and profoundly good.
For our budding AI sector, the elimination of non-competes would be a major win. To build on AI innovative momentum and preserve national competitiveness, Congress should seriously consider banning non-competes and backing the FTC’s action with legislative heft.
Labor Mobility and Dynamism
Why exactly should we act, and why are non-competes so bad?
First, dynamism. The AI industry thrives on mobility. In Silicon Valley where non-competes are illegal, for instance, labor is in constant flux allowing engineer supply to easily meet innovation demands.1 Clearly this worked. As an exceptionally dynamic field, AI requires a workforce capable of quickly pivoting, bringing their skills to promising new ventures, and applying their highly specialized knowledge to support the most innovative ideas. These agreements, however, hinder this essential fluidity. One 2009 study on Michigan’s legalization of non-competes, found a resulting 15% decrease in labor mobility among those with specialized technology skills. That is significant. Non competes put the brakes on job hopping, and in the process significantly hinder labor’s ability to shift on a dime to meet industry demand.
For those looking to start a business, these brakes pump even harder. When Hawaii banned tech industry non-competes in 2015, an Economic Innovation Group analysis found an 10.2% surge in business formation in just the following two years. Just by simply banning this type of contract, Hawaii unleashed an instant burst of pent up economic dynamism. While there are already a thousand flowers blooming in the AI industry, consider just how many more business and ideas could flourish without this unnatural constraint. Perhaps Hawaii’s business boom can be felt nationwide, across all industries.
Wasted Skill
A second risk is wasted labor. The severe shortage of qualified AI professionals is unquestionable. BCG reports that in 2024 AI-relevant sectors including electrical and software engineering will need roughly 100,000+ additional personnel. The result for AI: without talent, industry is curtailing investment and adoption and cutting-edge research halted. These challenges are compounded by non-competes. A 2011 survey of IEEE members (an engineering professional association which covers the industry) found that more than a third of members covered by non-competes have been forced to take needless “career detours.” Locked out of industry, talent is unnaturally pushed into menial, unrelated jobs, wasting critical skills and letting knowledge decay. 50% of engineers surveyed report signing a noncompete, if thirty percent of that fifty are forced out of the market that will put a substantial weight on industry. Momentum, competitiveness, and innovation will be lost.
Brain Drain
In the AI-specific case, an important consideration is international brain drain. According to Macro Polo 63% of top AI talent at American institutions are foreign born. This highly international workforce dependent on visas already has exceedingly low career mobility; restricted by the bureaucracy of visa transfers and the inability to go jobless for more than 60 days. Non-competes exacerbate this inflexibility. If blocked from inter-industry job transfers and restricted from taking ‘detours’ to time out contracts, visa holders can easily be forced, or lured, to their home country.
Among the best, brightest, and innovative, this is particularly likely. One 2014 study on interstate brain drain found the legalization of non-competes caused a 37% increase in the likelihood a given inventor (defined as a patent holder) moved to another state. Measuring the most professionally connected inventors, this impact rose to an eye-popping 236.3%.
This should be no surprise. Non-competes fundamentally stand between inventors and seeing their ideas flourish. Amid the dizzying financial promise of AI success, elite engineers will go wherever they need to reap the rewards of their labor. So be it if that means leaving the United States.
This effect is a needless brace on U.S. competitive success. The national security minded should note that that 38% of our workforce are Chinese born – a plurality over the 37% born domestically. These immigrants are critical to our success and should be encouraged to stay. They have a lot to offer. With inflexible policies, however, we encourage the drain of this talent’s ideas, innovations, and skills direct to our biggest geo-political competitor.
Potential risks?
The arguments in favor of non-competes largely surround IP theft protections. If employees can jump ship on a dime, they can easily leak memorized strategies, algorithms, and other critical information. Tyler Cowen notes a theoretical added effect is to encourage information siloing inside businesses and potentially constraining collaboration.
To these ends, it’s important to note that the law already provides a solution: IP theft is illegal. State and corporate resources are already being devoted to enforcement. If that’s not good enough, perhaps they should devote more. Alternatively placing the enforcement burden on employees through non-competes simply acts to externalize and spread the costs to people who often have no stake in the IP game.
In the AI case, it’s also questionable whether non-competes can halt theft and diffusion of secrets. As I’ve argued, algorithmic secrets are easy to memorize and easy to share. Non-competes cannot stop memorized algorithms from being anonymously posted online or taken abroad as talent is poached by foreign companies looking for advantage. That only grows more likely if employees are restricted domestically. Innovation diffusion is almost inevitable and the government, and companies, should instead lean into this assumption. Firms should focus less on viciously protecting secrets and information siloing and more on putting those secrets into use. If a firm shores up a solid product and user base, when secrets are inevitably leaked or stolen, their position will remain steady.
Naturally, there are many more points to make than this. For a comprehensive overview of the literature and whether other key justifications for non-competes hold water I highly recommend checking out this piece from Evan Starr.
Enabling Talent
When I worked in healthcare tech, I felt and saw firsthand the negative impacts of non-competes. Many friends leaving our company often wanted to stay in industry but found their careers put on hold for years by their former bosses. Many ended up jumping ship and biding their time in tangential fields, waiting for contracts to expire. The toll was incredible. Each sacrificed income, career momentum, and time; flourishing hindered by an uninvolved third party. Meanwhile the industry’s chronic labor shortage continued.
AI success and AI diffusion is ultimately driven by talent, and we should do our upmost to enable that talent to flourish. Forcing engineers to languish helps no one. While there are certainly bigger talent reform fish to fry (namely high skilled immigration reform), banning these agreements represents something perhaps politically feasible. If the FTC’s rule fails, Congress should push forward. Already, a majority of states have limited non-competes, and many are even starting to ban them. This has support; this has momentum. If Congress wants to build a world-leading AI industry and maximize our limited engineering talent base, they should swim with this forward current of political energy and legislatively ban non-compete agreements.
Special thanks to Phalen Kuckuck for editing and review.
Note that while Silicon Valley dominates the industry, AI goes well beyond the Bay Area. There are countless businesses across the nation developing models, countless remote workers push from California by staggering housing, and countless companies trying to fine-tune and tune AI systems to their bespoke industrial needs. I personally worked for a tech company leading innovation healthcare AI based outside of California. I - as with everyone there - had to sign a non-compete. Reform in California helps but doesn’t capture everyone. To promote maximum benefit, we need to capture everyone.