Anthropic Is Reportedly Asking Candidates: What If Safety Makes Your Equity Worthless?

Most interview questions about values are free to answer. According to people familiar with the process, Anthropic has added one with a price tag attached: how would you feel if the company decided, for safety reasons, to slow its AI ambitions and your equity went to zero?

According to Axios, whose reporting a Spanish-language Xataka piece reconstructed for its audience, Dario Amodei suspects some people are joining for the wrong reason, and the question is now part of the company's "culture interview," a conversation with a specific employee rather than someone from HR. One candidate, speaking anonymously, recalled it asked flat out: what happens if your shares are worth nothing because the company refuses to build something dangerous? The phrasing varies. Some hear the share price literally at zero, others a softer dilemma between safety and revenue.

On Blind, someone wrote that they answered honestly: they would not be happy, they want to do things right but also want to work at a business that works. The interviewer, reportedly, did not love it.

Cheap to say, expensive to mean

The question is sharper than it looks because of when it is being asked. Anthropic raised its valuation to roughly $965 billion in May, confidentially filed IPO paperwork in June, and is now discussed as a possible record-setting market debut. Candidates are being hired just before the bell, and the company is asking who would stay if the bell never rang.

The Xataka piece adds useful context. The culture interview has existed since the company's first days, per Daniela Amodei, but the zero-equity framing is recent, with mentions on Blind starting in 2025 and two former employees telling Axios it was not asked that way before. Meanwhile the company sells itself as the lab capable of saying no: it rejected a US Army contract that came with unrestricted use, and limited access to its most cybersecurity-capable model.

There is also a compensation angle. Anthropic pays well but, by all accounts, does not negotiate. In a sector where strong researchers command bidding wars and some candidates spend thousands on interview coaches to land offers worth hundreds of thousands more, a no-haggle policy plus a values interrogation is a distinctive filter. It selects for people who want in despite the money, or at least can credibly perform that.

The honest reading, both directions

Two things can be true at once.

The question is legitimate. Companies that claim a safety mission and then staff up with people purely chasing the exit are building a contradiction into their own workforce. The moment a competitor doubles your liquid offer, the mission evaporates. Anthropic is forcing that conversation before signing, not after. Plenty of companies should be having it.

And the question is also unverifiable theater until the shares actually trade. As Xataka notes, while the stock stays private, the "sacrifice" cannot be priced, and saying yes in an interview costs nothing. The people who spent heavily on coaching to pass the culture screen have already answered, with their credit card, whether they come for the mission or the salary. The answer that scores well is the mission. The honest answer, judging by Blind, is not always that one. One more wrinkle from the reporting: Amodei has pledged to give away most of his fortune, which makes the boss's own answer to the question easier to give than a mid-level engineer's.

Why this matters outside Anthropic

You can file this under labor-market trivia and move on. But the underlying question generalizes to anyone building with, on, or around generative AI right now.

If you are joining an AI company, ask yourself the zero-equity version: would the day job still be worth it if the windfall part went away? The honest answer tells you whether you are choosing a workplace or a lottery ticket. Both are valid choices, but knowing which one you made changes how you should evaluate the offer, the equity, and the culture questions they ask you.

If you are building a product on top of these labs, the interview question is a signal about your suppliers. A lab that seriously prices its own safety decisions into hiring is a lab whose roadmap can pause, change, or restrict access for reasons that have nothing to do with your release plan. Anyone whose workflow depends on a single provider's most capable model has already felt what a policy decision does to a product roadmap. Treat provider concentration as an engineering risk, not a political opinion, and keep a second model wired in.

And if you are building your own thing, the cheap-talk problem is worth stealing. Interview questions, mission statements, and team values are all free for candidates and companies alike. The only tests that mean anything are the ones with a real cost attached: giving up a contract, capping a feature, or pricing in the downside before you sign people up. Anthropic's question is uncomfortable precisely because it tries to attach a number to a value. Most companies never do.

The question is easy to pass and hard to live. That gap, between the answer that gets you hired and the answer you would actually act on, is the whole point. The same gap exists in every team building with these tools, and closing it before the pressure arrives is considerably cheaper than after.