OpenAI has recently made a significant move to rectify a controversial practice that has sparked considerable backlash. The artificial intelligence giant has decided to release its employees from a stringent non-disparagement agreement, an announcement that was made through a staff-wide memo. This move comes as a response to the most severe scandal faced by CEO Sam Altman during his tenure at the company.
The Memo That Changed Everything
In a memo circulated to both current and former employees, OpenAI clarified that it would no longer enforce the non-disparagement contract that employees had been pressured to sign since at least 2019. This agreement demanded that employees refrain from criticizing the company or risk losing their vested equity. For many, this could mean forfeiting millions of dollars for expressing even mild criticism of OpenAI’s operations.
The full memo was shared publicly by Andrew Carr, a former OpenAI employee, on X (formerly Twitter). Carr, who left the company in 2021, posted the memo to shed light on the restrictive practices that had governed employee departures.
High-Profile Resignations and Public Outcry
The scandal came to the forefront when Vox exposed the harsh terms of the non-disparagement clause. This revelation followed the unexpected resignations of OpenAI co-founder and chief scientist Ilya Sutskever and his superalignment team co-leader Jan Leike. Their abrupt departures led to rampant speculation about potential internal conflicts and unsafe AI practices within OpenAI.
Sutskever and Leike offered minimal explanations for their resignations, fueling further curiosity. Sutskever merely expressed his optimism about OpenAI’s future in building safe and beneficial artificial general intelligence (AGI), while Leike’s statement was simply, “I resigned.”
As public scrutiny intensified, many speculated that the non-disparagement agreements were gagging former employees from disclosing critical information about the company’s inner workings.
Financial Coercion and Uncommon Practices
According to Vox, employees were compelled to sign the exit agreements within a week of leaving the company, under threat of losing their vested equity. These agreements were described as “fairly uncommon in Silicon Valley,” allowing OpenAI to censor former employees indefinitely.
In response to the backlash, Altman took to X, admitting his embarrassment over the situation. He claimed ignorance of the practice, stating, “I did not know this was happening and I should have.” However, Vox reporter Kelsey Piper noted that Altman had recently signed separation letters containing the controversial terms, casting doubt on his claim of unawareness.
The controversy highlighted OpenAI’s use of financial pressure to expedite the signing of exit agreements and even threatened to block employees from selling their equity. Requests for additional time to review the agreements, which would allow employees to seek legal counsel, were reportedly denied as recently as this spring.
A Change in Policy
OpenAI’s Chief Strategy Officer, Jason Kwon, told Vox that the company had been reconsidering the language of these agreements for about a month before the scandal broke. He expressed regret over the distress caused to former employees and assured that the company was working swiftly to rectify the situation.
In a significant policy shift, OpenAI declared that it would never revoke vested equity from employees who left the company and subsequently criticized it, provided they did not sign a non-disparagement agreement. This assurance aimed to alleviate fears among former employees about financial retribution for expressing their true opinions about the company.
Acknowledging Past Mistakes
One former employee, Daniel Kokotajlo, publicly shared his decision not to sign the exit agreement, despite the significant financial risk involved. He emphasized the importance of retaining his ability to critique the company in the future. Kokotajlo revealed that his vested equity likely represented a substantial portion of his family’s net worth but felt it was crucial to maintain his right to speak freely.
OpenAI has since confirmed that it never enforced these non-disparagement agreements. Moving forward, the company announced that it would remove such clauses from its standard departure paperwork and release former employees from existing non-disparagement obligations, except in cases where the provision was mutual.
A Commitment to Transparency
In a memo to employees, OpenAI reiterated its commitment to not cancel any vested equity units, regardless of whether the departure agreement was signed. The company expressed deep regret for the delay in changing the language, acknowledging that it did not reflect the values or the company culture OpenAI aspires to embody.
“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,” said an OpenAI spokesperson, marking a pivotal moment in the company’s ongoing efforts to align its practices with its stated principles of transparency and fairness.
See also: ZeroPoint: Revolutionizing Memory Compression For AI Infrastructure
