Leaked OpenAI documents reveal aggressive tactics toward former employees

On Friday, Vox reported that employees at tech giant OpenAI who wanted to leave the company were confronted with expansive and highly restrictive exit documents. If they refused to sign in relatively short order, they were reportedly threatened with the loss of their vested equity in the company — a severe provision that’s fairly uncommon in Silicon Valley. The policy had the effect of forcing ex-employees to choose between giving up what could be millions of dollars they had already earned or agreeing not to criticize the company, with no end date.

According to sources inside the company, the news caused a firestorm within OpenAI, a private company that is currently valued at some $80 billion. As with many Silicon Valley startups, employees at OpenAI often get the majority of their overall expected compensation in the form of equity. They tend to assume that once it has “vested,” according to the schedule laid out in their contract, it is theirs and cannot be taken back, any more than a company would claw back salary that has been paid out.

A day after the Vox piece, CEO Sam Altman posted an apology, saying: 

we have never clawed back anyone’s vested equity, nor will we do that if people do not sign a separation agreement (or don’t agree to a non-disparagement agreement). vested equity is vested equity, full stop.

there was a provision about potential equity cancellation in our previous exit docs; although we never clawed anything back, it should never have been something we had in any documents or communication. this is on me and one of the few times i’ve been genuinely embarrassed running openai; i did not know this was happening and i should have.

Tl;dr: I didn’t know we had provisions that threatened equity, and I promise we won’t do that anymore.

That apology has been echoed in internal communications by some members of OpenAI’s executive team. In a message to employees that was leaked to Vox, OpenAI chief strategy officer Jason Kwon acknowledged that the provision had been in place since 2019 but that “The team did catch this ~month ago. The fact that it went this long before the catch is on me.”

But there’s a problem with those apologies from company leadership. Company documents obtained by Vox with signatures from Altman and Kwon complicate their claim that the clawback provisions were something they hadn’t known about. A separation letter on the termination documents, which you can read embedded below, says in plain language, “If you have any vested Units … you are required to sign a release of claims agreement within 60 days in order to retain such Units.” It is signed by Kwon, along with OpenAI VP of people Diane Yoon (who departed OpenAI recently). The secret ultra-restrictive NDA, signed for only the “consideration” of already vested equity, is signed by COO Brad Lightcap.

Meanwhile, according to documents provided to Vox by ex-employees, the incorporation documents for the holding company that handles equity in OpenAI contains multiple passages with language that gives the company near-arbitrary authority to claw back equity from former employees or — just as importantly — block them from selling it. 

Those incorporation documents were signed on April 10, 2023, by Sam Altman in his capacity as CEO of OpenAI. 

Vox asked OpenAI if they could provide context on whether and how these clauses made it into the incorporation documents without Altman’s knowledge. While that question was not directly answered, Kwon said in a statement to Vox, “We are sorry for the distress this has caused great people who have worked hard for us. We have been working to fix this as quickly as possible. We will work even harder to be better.” 

The seeming contradiction between OpenAI leadership’s recent statements and these documents has ramifications that go far beyond money. OpenAI is arguably the most influential, and certainly the most visible, company in artificial intelligence today, one that has the stated ambition to “ensure that artificial general intelligence benefits all of humanity.” 

A little more than a week ago, OpenAI executives were on stage introducing the company’s latest model, ChatGPT-4o, which they were proud to note was capable of carrying out highly realistic conversations with users (with a voice, as it turned out, that was a bit too close to that of actress Scarlett Johansson).

But bringing artificial general intelligence to the world is a role that demands enormous public trust and serious transparency. If OpenAI’s own employees haven’t felt free to voice criticism without risking financial retribution, how can the company and its CEO possibly be worthy of that trust?

(Vox reviewed many documents in the course of reporting this story. Key documents of public interest are reproduced below.)

High-pressure tactics at OpenAI

Throughout the hundreds of pages of documents leaked to Vox, a pattern emerges. Getting ex-employees to sign the ultra-restrictive nondisparagement and nondisclosure agreement involved threatening to cancel their equity — but it also involved much more. 

In two cases Vox reviewed, the lengthy, complex termination documents OpenAI sent out expired after seven days. That meant the former employees had a week to decide whether to accept OpenAI’s muzzle or risk forfeiting what could be millions of dollars — a tight timeline for a decision of that magnitude, and one that left little time to find outside counsel. 

When ex-employees asked for more time to seek legal aid and review the documents, they faced significant pushback from OpenAI. “The General Release and Separation Agreement requires your signature within 7 days,” a representative told one employee in an email this spring when the employee asked for another week to review the complex documents. 

“We want to make sure you understand that if you don’t sign, it could impact your equity. That’s true for everyone, and we’re just doing things by the book,” an OpenAI representative emailed a second employee who had asked for two more weeks to review the agreement. 

(I spoke with four experts in employment and labor law for perspective on whether the termination agreement and surrounding conduct was indeed “by the book” or standard in the industry. “For a company to threaten to claw back already-vested equity is egregious and unusual,” California employment law attorney Chambord Benton-Hayes told me in an emailed statement.)

Most ex-employees folded under the pressure. For those who persisted, the company pulled out another tool in what one former employee called the “legal retaliation toolbox” he encountered on leaving the company. When he declined to sign the first termination agreement sent to him and sought legal counsel, the company changed tactics. Rather than saying they could cancel his equity if he refused to sign the agreement, they said he could be prevented from selling his equity. 

The later documents the company sent him, which Vox has reviewed, say, “If you have any vested Units and you do not sign the exit documents, including the General Release, as required by company policy, it is important to understand that, among other things, you will not be eligible to participate in future tender events or other liquidity opportunities that we may sponsor or facilitate as a private company.” In other words, sign or give up the chance to sell your equity.

How OpenAI played hardball 

To make sense of that — and to see why it makes OpenAI’s recent apology so hollow — you need to understand what equity at OpenAI means. 

In a publicly traded company, like Google, equity just means shares of stock. Employees are paid partially in their salary and partially in Google stock, which they can hold or sell on the stock market like any shareholder.

In a private company like OpenAI, employees are still awarded ownership shares of the company (or, more frequently, options to purchase ownership shares of the company at low prices) but have to wait until an opportunity to sell those shares — which may not come for years. Large private companies sometimes do “tender offers” where employees and former employees can sell their equity. OpenAI hosts tender offers sometimes, but the exact details are a tightly kept secret. 

By announcing that someone who doesn’t sign the restrictive agreement is locked out of all future tender offers, OpenAI effectively makes that equity, valued at millions of dollars, conditional on the employee signing the agreement — while still truthfully saying that they technically haven’t clawed back anyone’s vested equity, as Altman claimed in his tweet on May 18. 

Vox reached out to OpenAI to clarify whether OpenAI has used or plans to use this tactic to cut former employees off from equity. An OpenAI spokesperson said, “Historically, former employees have been eligible to sell at the same price regardless of where they work; we don’t expect that to change.” It is not clear who authorized telling a former employee that he would be excluded from all future tender offers unless he signed.

And the ex-employees I spoke with were nervous that, whatever public reassurances the company may be making, the incorporation documents generally gave OpenAI many avenues for legal retaliation, making it less reassuring for the company to retreat from any specific one. 

In addition to clauses stating that vested equity will vanish if a former employee does not sign a general release within 60 days, the incorporation documents also contain clauses stating that, “at the sole and absolute discretion of the company,” any employee who is terminated by the company can have their vested equity holdings reduced to zero. There are also clauses stating that the company has absolute discretion over which employees are allowed to participate in tender offers in which their equity is sold.

“[Those] documents are supposed to be putting the mission of building safe and beneficial AGI first but instead they set up multiple ways to retaliate against departing employees who speak in any way that criticizes the company,” a source close to the company told me. 

These documents are signed by Sam Altman. OpenAI did not respond to a question about whether there was a contradiction between Altman’s public statements that he was unaware company documents included language about clawing back equity and the presence of these clauses in incorporation documents with his signature on them.

Why it matters

OpenAI has long positioned itself as a company that ought to be held to a higher standard. It claimed that its unique corporate structure — which involved a for-profit company governed by a nonprofit — would let them bring transformative technology to the world and ensure it “benefits all of humanity,” as the company mission statement reads, and not just the shareholders. OpenAI’s senior leadership has talked at length about their responsibilities for accountability, transparency, and democratic input, with Altman himself telling Congress last year that “my worst fears are that we — the field, the technology, the industry — cause significant harm to the world.”

But for all the high-minded idealism, OpenAI has also had its share of scandals. In November, Altman was fired by the OpenAI board, which said in a statement only that Altman “was not consistently candid with the board.” The clumsy firing provoked an immediate outcry from employees, especially as the board failed to provide any more detailed explanation of what had justified firing the CEO of a world-leading tech company. 

Altman soon arranged a deal to effectively take the company and most of its employees with him to Microsoft, before he was ultimately reinstated, with much of the board then resigning. 

At the time, the board’s language — “not consistently candid” — was puzzling. (Has anyone ever met a CEO who is consistently candid?) But six months on, it seems like we might be starting to see publicly some of the issues that drove the unexpected board conflagration. 

OpenAI can still set things right, and may now be getting started on the long and difficult process of doing so. They have taken some first, necessary steps. Altman’s initial statement was criticized for doing too little to make things right for former employees, but in an emailed statement, OpenAI told me that “we are identifying and reaching out to former employees who signed a standard exit agreement to make it clear that OpenAI has not and will not cancel their vested equity and releases them from nondisparagement obligations” — which goes much further toward fixing their mistake.

In a fuller statement, OpenAI said:

“As we shared with employees today, we are making important updates to our departure process. We have not and never will take away vested equity, even when people didn’t sign the departure documents. We’re removing nondisparagement clauses from our standard departure paperwork, and we’re releasing former employees from existing nondisparagement obligations unless the nondisparagement provision was mutual. We’ll communicate this message to former employees. We’re incredibly sorry that we’re only changing this language now; it doesn’t reflect our values or the company we want to be.”

I think that represents a huge step forward over the company’s initial May 18 apology; it is specific about the steps OpenAI is taking and involves proactively reaching out to former employees. But I think OpenAI’s work here is far from done. Former employees felt the company put them under pressure from multiple angles, and OpenAI has not yet committed to changing all of those — specifically, they should commit to not excluding anyone from selling their equity on the basis of not signing a document or criticizing Open AI. 

And, to fully grapple with the situation, OpenAI needs to grapple with responsibility. It’s hard to understand how the executive team could have signed documents that laid out avenues to claw back equity from former employees, as well as separation letters which threatened to do the same, without realizing this situation was happening. In order to set this issue right, OpenAI must first acknowledge how extensive it was.

Check out the documents supporting this reporting below:

Update, May 22, 7:32 pm ET: This story has been updated to include a fuller statement from OpenAI.