AI · Nov 9, 2026
OpenAI’s Control Room puts a fail-closed kill switch inside the inference engineAnthropic’s safety promise now rests on a trust with little real power
· Updated · Signal Desk newsroom
Analysis of reporting by The Verge. How we work

The Verge picks apart the Long-Term Benefit Trust as Anthropic completes the largest IPO ever. The structure insulates founders more than it checks them, which makes the company’s character the only real safeguard.
Key takeaways
- The Verge reports Anthropic’s IPO raised $81.3 billion and beat SpaceX as the most valuable ever, at a fully diluted market cap of about $1.5 trillion.
- The Long-Term Benefit Trust picks four of seven directors, but one needs effective founder approval and nominations route through a founder-chaired committee.
- Per the S-1, the trust mainly appoints and removes directors and lacks the day-to-day or release-blocking authority OpenAI’s nonprofit kept for safety.
- Governance experts, a watchdog and a former employee told The Verge they see little meaningful check on the founders.
- Dario Amodei called compute the company’s deepest business dilemma; public-market pressure will add to it.
A record raise, and a governance question
Anthropic is now a public company, and by The Verge’s count its offering raised $81.3 billion, the most valuable IPO ever. Our coverage of the Nasdaq debut showed shares opening at $111, 33% above the $83 price. The Verge’s Hayden Field asks the question that should follow any record: who can tell this company no?
Anthropic’s answer has long been its structure. It is a Public Benefit Corporation, so directors must weigh mission against shareholder returns, and its Long-Term Benefit Trust, created in 2023, holds a special Class T stock that lets financially disinterested trustees select and remove part of the board. On paper that sounds like a safety brake. The Verge’s reporting suggests it is closer to a steering committee with limited reach.
Reading the fine print
The trust technically chooses four of seven board seats, but one of those effectively needs founder approval, which leaves it short of clean majority control. Any nomination or removal requires 30 days’ notice to a three-person governance committee chaired by co-founder Daniela Amodei, alongside Reed Hastings and investor Yasmin Razavi. The trustees, including Ben Bernanke and chair Neil Buddy Shah, are distinguished but not known as AI experts, a point law professor Ann Lipton pressed directly. And nothing stops the company from hiring a trustee away; a former trustee, Mariano-Florentino Cuéllar, joined in August as chief global affairs officer.
The contrast with OpenAI is the most important detail. OpenAI’s nonprofit board kept authority over safety and security matters, such as delaying or stopping a model release, even after restructuring. Anthropic’s S-1 suggests its trust has neither that power nor day-to-day authority. Combined with the founder voting control we described when the filing first surfaced, the result resembles the founder-centric setups at Meta and SpaceX, not a novel check.
Where that leaves the safety brand
We do not read this as proof that Anthropic will abandon safety. Its leadership has argued consistently that safety requires being at the frontier, and it is unusually candid about risk, down to warnings in its own prospectus. But candor is not constraint. Lipton’s summary is the cleanest: investors either trust the founders to balance society’s interests or they don’t, and the binding limit will be the need for money. Amodei himself called compute the company’s deepest dilemma, too little stalls research, too much overextends it.
Critics are already pressing. The Verge cites a Center for AI Safety statement accusing Anthropic of weakening safety pledges and using safety branding to court investors, plus departures by safety researchers warning of escalating risk. Those are allegations and opinions, not findings. Still, the burden has shifted. A public Anthropic will be judged on observable choices, delayed releases, published evaluations, external audits, rather than on a governance chart that, by the evidence here, does not bind it much.
Who carries the risk
- Public shareholders
- They are buying into a company where founders and a limited trust, not shareholders, hold real governance power.
- Safety researchers
- Internal advocates lose leverage if the trust cannot block or delay a release.
- Regulators
- If self-governance is weak, the case for external oversight of frontier labs gets stronger.
- Rival labs
- Anthropic’s structure will become the template, or the cautionary tale, for the next AI IPOs.
What to watch
- Any change to trustee membership or a trustee moving into an Anthropic job.
- Whether Anthropic publishes safety evaluations that delay or alter a major model release.
- Shareholder or regulator pressure on the Class T arrangement.
- How Anthropic discusses recursive self-improvement now that quarterly results are public.
FAQ
What is the Long-Term Benefit Trust?
A body of financially disinterested trustees, created in 2023, whose Class T stock lets it select and remove some Anthropic directors.
Can the trust stop a model release?
According to The Verge’s reading of the S-1, no. Its main power is appointing and removing certain directors, unlike OpenAI’s nonprofit board, which retains safety authority.
How much did the IPO raise?
The Verge reports $81.3 billion; Signal Desk’s earlier coverage noted an $83 IPO price and a $111 opening trade.
More on OpenAI
AI · Nov 11, 2026
OpenAI’s $2 million for children’s AI safety research matters for one promise: publishing bad newsAI · Nov 17, 2026
OpenAI says its models now shave area off production chips, including AMD’s next flagshipsAI · Nov 3, 2026
OpenAI’s GPT-5.6 Cyber reaches AWS GovCloud through Daybreak Red, with two approval gatesAI · Oct 29, 2026
OpenAI’s new Model Spec swaps rules for five values, trading auditability for judgment