PBC’s approach to its planned initial public offering threatens to disempower shareholders and prevent ordinary people from reining in a potentially dangerous form of technology, according to New York City’s comptroller.

Details provided by Anthropic in its Oct. 26 IPO filing have raised a number of new concerns, Mark Levine said in an interview with Bloomberg. “We’re seeing a very worrisome trend of disempowerment to shareholders and all the balance of power tipping towards management.”Levine already made clear in mid-October that he thinks Anthropic should pause its IPO until it’s addressed the safety risks associated with artificial intelligence. In its S-1 filing, the company warned of the “catastrophic or existential risks to humanity” that AI may pose.

Anthropic pledged to slow down or even halt AI development if it thinks safety is at risk, but also acknowledged the technology may develop faster than humans’ ability to control it.

Levine, who as comptroller looks after New York’s fiscal health and oversees public pension funds with more than $325 billion in assets, said given such unprecedented risks, empowering regular shareholders is more important than ever. But the IPO filing makes clear that with Anthropic’s approach to governance, shareholder rights have been diminished, he said.

Levine’s criticism echoes some of the warnings that he and others leveled at Space Exploration Technologies Corp., which went public in June. The SpaceX IPO left Elon Musk with more than 80% of the voting rights, while also making him chief executive and chief technical officer, as well as chair of the board.

Concentration of control has been a trend in tech startups for years. “But what we saw in SpaceX, and now what we’re seeing in Anthropic, really, it’s at a whole new level,” Levine said. Spokespeople for Anthropic and SpaceX didn’t respond to requests for comment.

Levine said that, as is the case with SpaceX, the New York city pension funds he oversees will end up being exposed to Anthropic via indexes, despite his concerns.

Anthropic has designed its governance so that its seven co-founders and an independent trust retain control of the company after the IPO. The structure is supposed to allow Anthropic, which is listing as a public benefit corporation (PBC), to prioritize safety over short-term shareholder rewards. At the same time, the company’s board will set its own standards to determine what a successful public benefit mission looks like, according to its pre-IPO filing.

Anthropic also says it can’t assure stakeholders that it will achieve its corporate purpose or that the expected positive impact of being a PBC will be realized, its S-1 filing shows. Levine isn’t alone in questioning the governance model.

Though Anthropic doesn’t quite raise the same concerns as SpaceX, there are still material environmental, social and governance questions, said Miranda Beacham, head of UK responsible investment at Aegon Asset Management, which oversees close to $450 billion in assets. A key worry is the company’s stated need to continue tapping markets for large amounts of cash, which will impact existing shareholders in ways they ultimately can’t influence, she said.

Given the “huge capital requirements,” there’s a risk of “significant shareholder dilution,” Beacham said. “There are also legitimate questions about whether governance and safeguards can keep pace with increasingly powerful AI products and control how they are used.”She declined to comment on whether Aegon intends to buy shares in Anthropic.

Nell Minow, co-founder and chair of ValueEdge Advisors, a corporate governance and ESG consulting firm that advises institutional investors, asset managers and public pension funds, says that the structure adopted by Anthropic is one with which shareholders have had very little experience.

“I give them credit for their good intentions in trying to be different from SpaceX,” she said of Anthropic. “They are just making a completely new set of mistakes.”For some institutional investors, meanwhile, the governance concerns that for them rendered SpaceX uninvestable are largely absent from Anthropic.

The majority control that the founders get isn’t ideal, said Anders Schelde, the chief investment officer of Danish pension fund AkademikerPension. “But we’re also flexible because we think it’s a process and it’s understandable that the founders would like to retain some sort of control.”What’s more, the conduct and tone of public comments by Anthropic co-founder and chief executive, Dario Amodei, give investors some confidence that he’ll treat the risks ahead seriously, Schelde said.

Amodei and his team at Anthropic are clearly more genuinely concerned about safety than Musk is, Levine also said. “It’s much more part of the DNA of the company. And that is reassuring.”

But Anthropic has “in pretty significant ways accelerated” the AI race “in ways that potentially harm safety,” he said. And for that, the company “has to be held accountable.”