Compiled by the editorial desk with reference to Vox's investigation, public statements from OpenAI, and Daniel Kokotajlo's personal account.

OpenAI, the artificial intelligence research organization known for its namesake commitment to openness, is facing renewed scrutiny over the confidentiality terms it imposes on departing employees. A report by Vox, published this week, details how the company's off-boarding agreements include lifetime non-disparagement clauses, effectively barring former staff from any public criticism. The stakes are high: employees who refuse to sign, or who breach the terms, can forfeit all vested equity accrued during their tenure—a sum that, for senior researchers, can reach into the millions.

The report, authored by Kelsey Piper, cites a former employee, Daniel Kokotajlo, who publicly stated he left OpenAI in 2023 due to "losing confidence that it would behave responsibly around the time of AGI." Kokotajlo confirmed that he surrendered what would likely have been a substantial equity payout rather than sign the agreement. His case illustrates the practical consequences of the policy, which some legal experts describe as unusually restrictive even by Silicon Valley standards.

OpenAI responded to the Vox report with a statement asserting that it "has never canceled any current or former employee’s vested equity nor will we if people do not sign a release or nondisparagement agreement when they exit." When asked whether this represented a policy change, the company said only that the statement "reflects reality." The response does not directly address Kokotajlo's account, which suggests that the policy may have been applied inconsistently or that the company's public stance differs from its internal practice.

Why the NDA Matters for AI Governance

The controversy comes at a time when OpenAI and its competitors are under increasing pressure to demonstrate responsible AI development. The company has long positioned itself as a leader in safety and transparency, but the off-boarding terms appear to contradict that narrative. Critics argue that such agreements stifle whistleblowing and public accountability, particularly as AI systems become more powerful and their societal impacts more profound.

The use of non-disparagement clauses is not unique to OpenAI; many tech companies employ similar provisions to protect trade secrets and brand reputation. However, the lifetime duration and the linkage to vested equity are notable. Legal analysts note that while non-disparagement clauses are common, tying them to already-earned equity is a more aggressive tactic, as it creates a financial penalty for speech rather than merely a contractual obligation.

For Kokotajlo, the decision to forgo his equity was a matter of principle. In his public statements, he has expressed concerns about OpenAI's trajectory, particularly its focus on artificial general intelligence (AGI) and the potential risks associated with it. His departure is part of a broader pattern of high-profile exits from the company, including researchers who have cited safety concerns.

Industry Context and Future Outlook

The Vox report has reignited debate about the governance of AI development. Some observers argue that the industry's reliance on private funding and corporate incentives creates inherent conflicts between transparency and profitability. Publicly funded research institutions, they contend, might offer a more accountable path forward, but such a shift appears unlikely in the current economic climate.

OpenAI's denial, while categorical, leaves unresolved questions about the scope of its policy and whether it has been applied uniformly. The company's refusal to clarify whether the statement represents a change in practice suggests that the issue may persist. For now, the episode underscores the tension between the company's stated values and its contractual realities.

As the AI sector continues to expand, the treatment of departing employees will likely remain a focal point for journalists, regulators, and the public. The outcome of this controversy could influence how other firms structure their off-boarding agreements, and whether equity-based penalties for speech are deemed acceptable in the industry.