Anthropic going public could be a win for transparency. Instead, the early signs suggest the opposite: a public company that wants the money and prestige of the public markets, without the messy part where the public gets a real say.
That’s why I’m not brushing off the New York City comptroller’s criticism of Anthropic’s IPO plans as political noise. If anything, it’s the kind of friction we should expect when a company building powerful AI systems asks regular investors to buy in while also trying to lock down how it can be questioned, challenged, or steered.
Based on what’s been shared publicly, the comptroller is warning that Anthropic’s planned IPO could weaken shareholder power and limit public oversight of emerging tech. This comes after Anthropic filed its S-1 in October 2026. The company is a public benefit corporation, which means it’s set up to balance shareholder value with a stated public mission tied to safer AI.
That “balance” is the whole point. And it’s also the easiest thing in the world to abuse.
On paper, public benefit status sounds like the grown-up version of tech: not just “make money,” but “make money and do good.” In practice, it can turn into a shield. If leadership wants to ignore shareholders, they can say they’re protecting the mission. If they want to ignore the mission, they can say they have to protect shareholder value. It becomes a choose-your-own-excuse machine unless governance is designed to force real accountability.
The comptroller’s argument, as I read it, is basically: don’t let “benefit” become a loophole that keeps the public out. And yes, shareholders aren’t “the public.” But in a public company, shareholder rights are one of the few levers outsiders have that can actually move management. We can pretend otherwise, but most other forms of “oversight” are just vibes until something goes wrong.
And with AI, “something goes wrong” doesn’t have to mean sci-fi disaster. It can mean a model that quietly becomes the default tool inside hospitals, schools, call centers, and city agencies. It can mean a system that nudges hiring decisions in one direction, or spits out confident nonsense that people rely on because they’re overwhelmed. It can mean a company making behind-closed-doors choices about what it will and won’t build, and who gets access, with almost no way for outsiders to push back.
If Anthropic wants public-market capital, it should accept public-market scrutiny. That’s the deal. You don’t get to take the benefits and then complain that accountability is “distraction.”
Now, I can already hear the counterargument: AI companies need stability. They need room to make long-term safety calls without getting bullied by short-term investors. If shareholder empowerment means “maximize quarterly results,” then empowering shareholders could make AI less safe, not more safe. That’s a real concern. I’m not dismissing it.
But there’s a difference between protection from short-term pressure and protection from consequences. If an IPO is structured so regular shareholders can’t meaningfully influence governance, then what you’re really saying is: “Trust us forever.” And I’m sorry, but “trust us” is not a serious governance model for technology that could shape how millions of people work and live.
Imagine you’re a pension fund investing on behalf of city workers. You’re not buying a meme stock. You’re trying to protect retirements. If the company’s structure makes it hard to challenge leadership even when something looks off, you’re basically funding a black box with a mission statement taped to the side.
Or say you’re a regular person buying a small amount in an IPO because you believe in “safer AI.” Later, you learn the company made a major change in how it handles access or safeguards, and you disagree. What do you do? Sell? Complain online? That’s not oversight. That’s quitting.
And if you’re the company, notice what that creates: a feedback loop where the people most able to demand accountability are the least able to do it. Management gets more freedom, not less, right when the stakes get higher.
I also don’t love the quiet assumption that “public oversight” only matters after disaster. With AI, the real damage can be slow and boring. A slight drop in service quality that becomes normal. A workplace that becomes more automated and more monitored. A school district leaning on tools because they’re cheap, even if they’re wrong often enough to hurt kids who already struggle.
None of this proves Anthropic is planning something bad. The comptroller’s criticism is still just a warning, and a warning can be unfair. But I think the direction of travel matters. If the IPO is designed in a way that reduces shareholder voice and limits scrutiny, that’s not a neutral choice. That’s a statement: “We don’t want to be governed.”
And the thing about governance is you only notice it when you need it.
So here’s what I can’t shake: if Anthropic genuinely believes its public benefit mission is real and sturdy, why not welcome stronger shareholder rights and clearer oversight instead of treating them like a threat?