OpenAI and a new scandal

OpenAI has recently come under fire for lifelong anti-criticism policies included in severance agreements for departing employees.

The agreement includes non-disclosure and non-criticism provisions that prohibit former employees from criticizing OpenAI for the rest of their lives. Even acknowledging the existence of an NDA is a violation of the agreement. If a departing employee refuses to sign the document or violates it, he or she risks losing all the accumulated shares earned while working for the company, which could amount to millions of dollars.

Public reaction and legal issues:

The policy has been described as overly expansive and potentially unenforceable, with some commentators suggesting that any judge in California would rule the entire NDA unenforceable. Lawyers have pointed out that such agreements may not hold up in court, especially under California law, which has specific clauses that could invalidate such clauses.

Sam Altman's answer:

OpenAI CEO Sam Altman acknowledged the existence of a non-criticism clause, but said it was never enforced. He expressed regret over this and announced that the company would no longer include this clause in future exit agreements. Altman stressed that OpenAI has never “taken shares away from anyone” and assured that the company will not do this in the future.

Impact on employee departure:

The problem became apparent after the departure of prominent AI security researchers, including Jan Leike and Ilya Sutskever. Leike publicly criticized OpenAI's security culture, while Sutzkever remained tight-lipped about the reasons for his departure, possibly due to an agreement not to allow criticism. Former employee Daniel Kokotaylo confirmed that he had to give up a significant financial sum to leave without signing the document, highlighting the financial pressure brought by the agreement.

Widespread implications:

The dispute has raised questions about OpenAI's commitment to transparency and consistency with its stated mission of ensuring artificial general intelligence benefits all humanity. The restrictive nature of the agreements stands in stark contrast to the company's "open" name and mission, which has led to criticism from various quarters.