OpenAI Whistleblowers Allege NDAs Prevent Airing of Safety Concerns

Improve the News Foundation -
0:00
/1861

Facts

  • A letter from OpenAI whistleblowers to the US Securities and Exchange Commission (SEC) alleges that the ChatGPT-maker's use of non-disclosure agreements (NDAs) restricts them from sharing safety concerns with federal regulators.1
  • The letter, which was provided to news organizations by the office of Sen. Chuck Grassley (R-Iowa), claims that OpenAI had hires waive their rights to federal whistleblower compensation and that they required company consent to speak to regulators, with no exceptions for SEC violations.2
  • In the letter, the whistleblowers claim that the risks associated with AI development create an 'urgent need' for employees to be able to share concerns with regulators, which they say their employment and severance agreements prevented them from doing.3
  • The whistleblowers encourage the SEC to investigate every employment agreement at OpenAI and issue fines for each violation of federal law that allows employees of a company to share information with regulators without fear of retaliation.4
  • An OpenAI spokesperson defended the company's whistleblower process as one that 'protects employees' rights,' adding that the company encourages 'rigorous debate' and has removed non-disparagement clauses for departing employees.4

Sources: 1Guardian, 2Reuters, 3The Hill and 4Washington Post.

Narratives

  • Narrative A, as provided by Washington Post. The concerns raised by this letter are disturbing, to say the least. There are other reports trickling out of OpenAI that claim that the company hastily rushed through safety screening to meet a release date in May, as safety takes a backseat to financial interest. OpenAI cannot be trusted to fulfill its mandate of altruistic AI development and must be held to account.
  • Narrative B, as provided by reuters.com. No one has a better understanding of AI safety than the undisputed industry leader, and OpenAI has always made safety paramount. The company has a new safety committee working on a report on their upcoming model and already has strong whistleblower protections. There's no evidence that OpenAI is engaged in any nefarious activity or suppression of dissent.

Predictions