Anthropic is asking candidates whether they'd accept watching their equity go to zero if safety forced the company to stop building frontier AI.
According to a report from Axios, Anthropic has folded that question into a culture interview that every candidate goes through. The stage isn't a coding test. It isn't a standard HR screen either. Axios says the interview is run by employees nominated for the job, and one of its bluntest prompts asks candidates to imagine the company walking away from its AI work for safety reasons, taking the stock and their own equity down with it.
Would they support that call?
One candidate told Axios they answered honestly: no, they wouldn't be happy watching the stock go to zero, and they'd want to balance doing the most good with building a business that actually survives. The interviewer, the candidate said, didn't seem to like that answer.
That's a striking thing to select against. Most employers spend interviews trying to convince you the stock will go up. Anthropic is testing whether you'd still back the mission if the stock became worthless.
Anthropic's Dario Amodei Blames AI Backlash on a Crisis of Trust Anthropic CEO Dario Amodei says the AI industry's public backlash is "fundamentally a crisis of trust," pushing back on claims his own risk warnings are to blame. He admits AI firms haven't delivered on big promises like curing disease, even as lawsuits against rivals like xAI mount. - why people distrust AI companies right now - Anthropic CEO addresses artificial intelligence backlash crisis
Axios reports that two former Anthropic employees said this specific question wasn't part of the process in earlier hiring rounds. Daniela Amodei, Anthropic's co-founder and president, said in a 2024 Bain Capital Ventures interview that Anthropic has had a culture interview since day one. Both things can be true: the culture screen is old, but the stock-to-zero framing appears to be a newer and sharper version of it.
The money makes the question harder #
Look at the numbers and the question stops feeling abstract. Anthropic raised $13 billion in a Series F round on September 2, 2025, valuing the company at $183 billion, according to the company's own announcement. Then came more. On May 28, 2026, Anthropic said it had raised another $65 billion, at a $965 billion post-money valuation. Less than a week later, TechCrunch reported the company had confidentially filed for an IPO. That's quick.
The Financial Times reported this month that Anthropic is planning an October 2026 listing that could value it at about $2 trillion. The Week also noted that such a deal could become the largest IPO in history. That's the backdrop. A candidate is being asked whether they'd support a safety decision that wipes out their own paper wealth. This isn't a seminar question. It's a personal money question with a company badge on it.
CEO Dario Amodei has questioned whether newer employees are joining Anthropic for the right reasons, Axios reported. That's a fair worry at a lab whose valuation has raced from $183 billion to nearly $1 trillion in less than a year. But here's the thing: once you attach life-changing equity to a job, you can't be surprised when candidates think about the money. You invited that calculation into the room.
Anthropic's careers page leans hard on mission language, including its public benefit structure and its stated goal of making AI safe. It also says the company offers salary and equity packages, donation matching for equity grants, and regular office attendance for most staff. Those details matter because they show Anthropic isn't pretending compensation doesn't exist. It is trying to hire people who will treat compensation as secondary when the company says safety demands it.
This is not only interview theater #
Anthropic has already paid a price for safety positions. Axios reported in February and March 2026 that the company's dispute with the Pentagon centered on whether Claude could be used for fully autonomous weapons or mass surveillance of Americans. Axios also reported that the fight put tens of billions of dollars in direct and indirect government contracts at risk. That is real money.
The same pattern shows up in product access. Anthropic launched Claude Fable 5 and Claude Mythos 5 on June 9, 2026, then suspended access three days later. Its own launch post said Mythos 5 had the strongest cybersecurity capabilities of any model in the world and would be limited to Project Glasswing partners and other vetted defenders, with safeguards and access controls around risky cyber use. On August 21, the company said it was expanding Mythos 5 access through Claude Security and partner tools, while still keeping guardrails around direct model access.
The Trump administration designated Anthropic a national security supply-chain risk, forced it to recall its Mythos AI model under export controls, and banned federal agencies from using its products. While OpenAI and Google secured federal contracts by playing the Washington access game, Anthropic's refusal to allow autonomous weapons use has... - why did Trump ban Anthropic - Anthropic supply chain risk designation Pentagon
That makes the interview question less strange, though not less severe. Anthropic wants employees who will back costly restraint before they have a badge, a manager and a vesting schedule. Frankly, that's unusual. OpenAI and Google DeepMind ask candidates about values too, but Axios has not reported a matching practice in which they ask candidates to endorse the destruction of their own equity if mission and money collide.
There is a principled version of this. A frontier AI company that says it may need to slow down, restrict access or refuse lucrative contracts should hire people who understand that promise. There is also a performative version, where a company worth nearly $1 trillion asks workers to prove purity while founders and early investors sit on vastly larger stakes. You don't need to be cynical to see the tension.
Anthropic confirmed the existence of the culture interview to Axios but did not comment in detail on the specific process. That silence leaves the hardest question sitting where it belongs, with the company itself: whether the stock-to-zero test is a serious safety filter, or a way to ask ordinary employees for a level of sacrifice that will look very different after an IPO.
Also read: Thomson Reuters Built Its Own AI Model to Loosen Its Grip on Claude • Druckenmiller Dumps Intel, Micron and Broadcom for AMD and Bitcoin Miners • Google's Gemini 3.7 Flash Beats Rivals on Agent Benchmarks at Half the Price