OpenAI is facing renewed criticism from former employees over departure agreements that restricted criticism of the company and could expose employees to significant financial losses. The latest public comments come from Gretchen Kruger, who worked in policy research at OpenAI for almost five years, and Jacob Hilton, another former OpenAI employee who says he signed a gag order about a year ago.
Their statements add to an already visible debate about whether major AI labs give workers enough freedom to raise concerns in the public interest. The issue is not only about employment contracts. It is also about how organizations developing transformative AI handle accountability, transparency, and dissent from people who have worked inside them.
What the gag orders required
According to the source article, former OpenAI employees had for years been required to sign a document when leaving the company saying they would not criticize their former employer. If they did not comply, they could face significant financial losses.
OpenAI's CEO, Sam Altman, claimed to be unaware of these gag clauses. The source article also states that he had signed contracts that allowed the practice.
The practical effect of such a clause is clear from the accounts now being shared: former employees who might have concerns about OpenAI, its governance, or its handling of advanced AI work could feel pressure to stay silent. That pressure was not only reputational. In Hilton's case, he says it was tied to vested equity.
Gretchen Kruger expands the criticism
Gretchen Kruger worked in policy research at OpenAI for almost five years. She says she shares the concerns of the two safety researchers, Jan Leike and Ilya Sutskever, who recently left the company.
Kruger also points to a broader list of problems she believes need attention. She writes that fundamental improvements are needed in "decision-making processes; accountability; transparency; documentation; policy enforcement; the care with which we use our own technology; and mitigations for impacts on inequality, rights, and the environment,"
That list is notable because it moves beyond the narrow question of whether former employees should be allowed to criticize a company. It frames the issue as part of a wider governance challenge inside powerful technology organizations.
Kruger also warns that tech companies could weaken accountability efforts by dividing the people who want to hold them accountable. Her stated aim is to prevent that kind of division.
Daniel Kokotajlo and the role of disclosure
Kruger praises her colleague Daniel Kokotajlo, who previously refused to sign OpenAI's gag clauses and made them transparent to Vox. His refusal matters in this account because the existence and nature of the clauses became visible rather than staying hidden behind the same confidentiality expectations that employees were being asked to accept.
The source article presents Kokotajlo's action as an important step in making the issue public. Once the clauses were visible, other former employees could speak about the pressure they experienced and about why they believed the arrangement conflicted with the public responsibilities of a major AI lab.
Jacob Hilton says equity pressure shaped his choice
Jacob Hilton says he signed the gag order about a year ago. His explanation is direct: "I signed a non-disparagement a greement, with non-discl osure about the agreement itself, for no other reason than to avoid losing my vested equity," Hilton writes.
Hilton says he left OpenAI on a positive note and had no intention of criticizing the company. Even so, he says he was disappointed by the contract. In his view, it did not fit with a company that promises to develop artificial intelligence for the benefit of humanity.
His concern is not framed as a personal grievance. It is about the responsibilities of organizations building advanced AI. As Hilton writes, "Because of the transformative potential of AI, it is imperative for major labs developing advanced AI to provide protections for those who wish to speak out in the public interest,"
According to Hilton, OpenAI has since contacted him about lifting the confidentiality clauses.
Why this matters for AI accountability
The dispute highlights a central tension for major AI labs. These companies work on technology they describe as highly consequential, while also operating as employers with private contracts, internal policies, and financial incentives.
When former employees say they were restricted from criticizing a company, the public accountability question becomes more urgent. People who worked inside the organization may be among those best positioned to identify concerns about decision-making, documentation, policy enforcement, and the use of the company's own technology.
The accounts from Kruger and Hilton do not resolve the broader debate over OpenAI's practices. They do, however, add specific evidence that former employees saw the gag clauses as inconsistent with the level of openness and protection they believe advanced AI work requires.
For readers following AI governance, the key point is straightforward: the controversy is no longer only about one contract clause. It is about whether major AI labs can credibly claim public-interest missions while using agreements that discourage former employees from speaking openly about their concerns.