The (Not So Hidden) Reason that OpenAI is Hemorrhaging Talent ($$$)

OpenAI has lost (and keeps losing) many star players while Google, Anthropic, and other AI shops are catching up—and overtaking—OpenAI's lead. A small change in California's labor laws might be why.

Where there’s smoke, there’s fire.

I learned throughout my career that when top talent starts jumping ship, it’s time to head for the doors. The rockstars and linchpins in a company are the canary in the coalmine. When they lose faith in the business’s leadership, they are the first to go. Why? Usually, they have a central role and higher visibility. I call this the “burden of competence.” They are the prime movers in the company, so they see more than most. Furthermore, the fact that they are top talent means they are highly portable. Talent poaching is very real, and actually a deliberate function of the market. Neoliberal theory says that if a company is struggling, the market should punish that company, or even kill it. A mobile labor force is just one way that the market can create eternal churn and growth. Competition baby! Adapt or die!

The opposite is also true; it’s a good sign when dead weight is leaving a company. Here are a couple quick examples from my career.

Saul was a key player at a midsized MSP (managed service provider) I worked at. He had the Midas touch. When Saul showed up, everything went smoothly. The ship just didn’t run right without him around. So on the day that he showed up, fully erect posture, and a goofy grin plastered on his face, I knew something was up. This was the smile of liberation. He’d just gotten a job offer from the slick competitor, a league above us. My first thought was “Oh fuck, now what?” I should have started spamming my résumé that day. But I’m a slow learner.

Conversely, at another job, a midlevel senior manager started getting pissy with me. He was dumping work on me that was his responsibility, and just generally causing chaos. My boss intervened and told him to back off. Dave is mine. But Steve wasn’t done. He kept throwing wrenches in our gears, and I had a one-on-one call with another coworker, someone who’d been at the company more than 30 years. “Steve is a asshole!” He declared, fuming. I pointed out, “Steve is manipulative.” And my coworker lit up “Yes! That’s exactly right!” A year later, months after Steve quit the company, I’d learn that he’d been angling for my boss’s job, but was denied because he was a shifty twerp. When the bozos and clowns are leaving a company, that’s a sign that the culture is improving.

The fact that OpenAI has lost a huge amount of their most visible talent, to me, is a strong indicator that things are not all right in Eden. That alone should tell you something, but I suspect there’s another factor at play, and it has to do with simple rules shaping market dynamics.

California Revamps Non-Compete Bans

California has had a ban on non-compete agreements for a long time. However, in 2023, they introduced two new bills that revamp, reassert, and most importantly, expand the reach of their non-compete ban. Those two bills are Senate Bill 699 (SB 699) and Assembly Bill 1076 (AB 1076). Here’s a high level of what they do:

Senate Bill 699 (SB 699), which took effect on January 1, 2024, significantly strengthened California’s existing ban on noncompete agreements. It declared that any contract voided under California's noncompete prohibition is unenforceable regardless of where and when it was signed, effectively extending the ban’s reach to out-of-state agreements. SB 699 also established a private right of action for employees, allowing them to seek injunctive relief and actual damages, plus attorney’s fees and costs, if an employer attempts to enforce a void noncompete agreement.

As if that wasn’t spicy enough:

Assembly Bill 1076 (AB 1076), also effective from January 1, 2024, codified existing case law by explicitly making it unlawful to include noncompete clauses in employment contracts or require employees to enter noncompete agreements, unless they fall under specific statutory exceptions. Crucially, AB 1076 mandates that employers must notify current employees and former employees (employed after January 1, 2022) in writing by February 14, 2024, that any noncompete agreements or clauses they may have signed are void under California law. This notification must be sent to the employee’s last known postal and email addresses. Failure to provide this notice constitutes an act of unfair competition and may result in penalties of up to $2,500 per violation.

When you combine this with the embarrassment of OpenAI’s deeply problematic “clawback” clause coming to light, the net effect is that a lot of OpenAI employees were liberated from draconian laws and reminded that they had rights.

You know it’s embarrassing when the CEO has to tweet about it. You think he didn’t know?

Quite a few former OpenAI employees have moved onwards and upwards. This is to be expected—people learn and grow, and when they are qualified for a better job, they are liable to take one. In some cases, though, they’ve gone on to start rival companies. Anthropic, one of their chief rivals, was spawned by former OpenAI employees. Ilya Sutskever’s SSI, Inc. likewise came from a high level departure. In both cases, it was over a difference in vision. More recently, Mira Murati, whose pedigree includes Tesla, departed and is now building a new startup. All of these folks have attracted massive investment.

Plenty of other people, from lead researchers to safety specialists, have departed, citing a difference of opinion with respect to safety, ethics, and research directions. Some of that could be explained by OpenAI recently partnering with Anduril, a high-tech military/weapons company. Many people in tech have some high-minded values around the Military Industrial Complex, which is somewhat ironic to me. All of Silicon Valley serves as an appendage of the MIC. But I digress.

Labor Movement, Market Wins

Neoliberalism is an economic doctrine which places, at it’s heart, market primacy. Many capitalists mistakenly believe that we should favor business above all else. Others, such as UK’s Labour party, believe we should prioritize labor rights above all else. A true neoliberal, however, will recognize that neither side is right, not entirely. We should favor the market above all else. Sometimes that means taking rights from workers to rebalance the market, such as by dissolving unions. In other cases, it means taking rights away from businesses, such as by banning noncompete clauses.

Let’s examine the aggregate effect of all these high level departures from OpenAI to see the wisdom in emphasizing market primacy over business primacy or worker primacy.

1. Many new startups are forming, which is good for competition. Competition is good for everyone, in this case. The consumers win, the government wins, the economy wins. The brain drain from OpenAI represents a very steep gradient of talent reallocating itself across the market, allowing the market itself to explore more business paradigms, technical strategies, and market niches. Keeping talent locked up inside one company might have been better for that one company, but this alternative is better for everyone.

2. Top talent is circulating to other companies, such as Google, Meta, Microsoft, and more. The revolving door phenomenon is an open secret that in the tech world. People come and go, finding a place where their talent and values align better with companies. You get boomerang effects. This is good for workers, but it’s also good for companies engaging in talent poaching, they get to circulate that tribal knowledge, making the whole market more dynamic and productive.

3. Money and VC are flowing in like mad. When Mira left OpenAI, she had investors banging on her door immediately. Ilya quickly raised capital with a billion dollar valuation. There is always opportunity in chaos, and with latent talent and potential now liberated, America is in an even stronger position globally. Sure, OpenAI was the darling, but wouldn’t it better if we had a dozen OpenAI’s instead of just one?

What’s good for America is not always what’s best for an individual company. Likewise, what’s best for America is sometimes at odds with individual workers. I’m not expressly endorsing the system as-is, I’m merely characterizing how it works and why this is the current philosophy. I do personally celebrate every time someone leaves OpenAI, as I still feel betrayed by their decision to go closed-source and maximum-profit. I was an early adopter and advocate, and I think the market is rightly punishing them for Sam’s decision to upend his values and mission. The is-ought problem still applies here. I’m merely observing how the world is, while avoid prescriptions of how it ought to be. I have had my criticisms of both neoliberalism and capitalism, but I also try to remain grounded in reality. It’s a complex, messy world. It is far from ideal.

But at least in this one case, I see the absolute beauty of simple rules as high-leverage interventions which are potentially having outsized impacts on market dynamics.

Before we close out, I will say that there are certainly many reasons that people choose to leave companies. Value misalignments, better opportunities, burnout—I’m not trying to say that a pair of California bills are solely responsible. At the same time, I do appreciate the intention and apparent impact that these bills have had, and want to emphasize this as an example of a high-leverage intervention—a simple rule change that can have a butterfly effect on the fate of a nation.