Ex-OpenAI workers ask state AGs to block for-profit conversion

Former employees of OpenAI are asking the top law enforcement officers in California and Delaware to stop the company from shifting control of its artificial intelligence technology from a nonprofit charity to a for-profit business.

They’re concerned about what happens if the ChatGPT maker fulfills its ambition to build AI that outperforms humans, but is no longer accountable to its public mission to safeguard that technology from causing grievous harms.

“Ultimately, I’m worried about who owns and controls this technology once it’s created,” said Page Hedley, a former policy and ethics adviser at OpenAI, in an interview with the Associated Press.

Backed by three Nobel Prize winners and other advocates and experts, Hedley and nine other ex-OpenAI workers sent a letter this week to the two state attorneys general.

The coalition is asking California Attorney General Rob Bonta and Delaware Attorney General Kathy Jennings, both Democrats, to use their authority to protect OpenAI’s charitable purpose and block its planned restructuring. OpenAI is incorporated in Delaware and operates out of San Francisco.

OpenAI said in response that “any changes to our existing structure would be in service of ensuring the broader public can benefit from AI.” It said its for-profit will be a public benefit corporation, similar to other AI labs like Anthropic and tech billionaire Elon Musk’s xAI, except that OpenAI will still preserve a nonprofit arm.

“This structure will continue to ensure that as the for-profit succeeds and grows, so too does the nonprofit, enabling us to achieve the mission,” the company said in a statement.

The letter is the second petition to state officials this month. The last came from a group of labor leaders and nonprofits focused on protecting OpenAI’s billions of dollars of charitable assets.

Jennings said last fall she would “review any such transaction to ensure that the public’s interests are adequately protected.” Bonta’s office sought more information from OpenAI late last year but has said it can’t comment, even to confirm or deny if it is investigating.

OpenAI’s cofounders, including current CEO Sam Altman and Musk, originally started it as a nonprofit research laboratory on a mission to safely build what’s known as artificial general intelligence, or AGI, for humanity’s benefit. Nearly a decade later, OpenAI has reported its market value as $300 billion and counts 400 million weekly users of ChatGPT, its flagship product.

OpenAI already has a for-profit subsidiary but faces a number of challenges in converting its core governance structure. One is a lawsuit from Musk, who accuses the company and Altman of betraying the founding principles that led the Tesla CEO to invest in the charity.

While some of the signatories of this week’s letter support Musk’s lawsuit, Hedley said others are “understandably cynical” because Musk also runs his own rival AI company.

The signatories include two Nobel-winning economists, Oliver Hart and Joseph Stiglitz, as well as AI pioneers and computer scientists Geoffrey Hinton, who won last year’s Nobel Prize in physics, and Stuart Russell.

“I like OpenAI’s mission to ‘ensure that artificial general intelligence benefits all of humanity,’ and I would like them to execute that mission instead of enriching their investors,” Hinton said in a statement Wednesday. “I’m happy there is an effort to hold OpenAI to its mission that does not involve Elon Musk.”

Conflicts over OpenAI’s purpose have long simmered at the San Francisco institute, contributing to Musk quitting in 2018, Altman’s short-lived ouster in 2023 and other high-profile departures.

Hedley, a lawyer by training, worked for OpenAI in 2017 and 2018, a time when the nonprofit was still navigating the best ways to steward the technology it wanted to build. As recently as 2023, Altman said advanced AI held promise but also warned of extraordinary risks, from drastic accidents to societal disruptions.

In recent years, however, Hedley said he watched with concern as OpenAI, buoyed by the success of ChatGPT, was increasingly cutting corners on safety testing and rushing out new products to get ahead of business competitors.

“The costs of those decisions will continue to go up as the technology becomes more powerful,” he said. “I think that in the new structure that OpenAI wants, the incentives to rush to make those decisions will go up and there will no longer be anybody really who can tell them not to, tell them this is not OK.”

Software engineer Anish Tondwalkar, a former member of OpenAI’s technical team until last year, said an important assurance in OpenAI’s nonprofit charter is a “stop-and-assist clause” that directs OpenAI to stand down and help if another organization is nearing the achievement of better-than-human AI.

“If OpenAI is allowed to become a for-profit, these safeguards, and OpenAI’s duty to the public can vanish overnight,” Tondwalkar said in a statement Wednesday.

Another former worker who signed the letter puts it more bluntly.

“OpenAI may one day build technology that could get us all killed,” said Nisan Stiennon, an AI engineer who worked at OpenAI from 2018 to 2020. “It is to OpenAI’s credit that it’s controlled by a nonprofit with a duty to humanity. This duty precludes giving up that control.”