What is different about interviewing for Radar at Stripe?
Radar is fraud and risk, and the new work is not another card-testing story. Sessions 2026 expanded Radar toward token theft at AI companies, and the seat includes a pricing narrative for offerings that are still early. You should be able to write one fraud rule in plain language and explain who it protects. A generic “trust and safety” answer will not survive the follow-up.
Read this role on Stripe’s site
Token theft changed what Radar has to explain
Card testing is the story everyone brings. Radar’s 2026 expansion is token theft at AI companies, and that buyer is not a store manager.
An AI company loses money when someone steals tokens, not when a script tries a thousand cards. The product still has to be explainable to a support lead and a finance partner who already bought a fraud tool. Your job is the difference, in their words.
Zero-to-one work sits on this seat, including the price. Be ready to say what you would charge for, what you would leave free, and which metric would tell you the price was wrong.
Cases that only make sense if you own Radar
They will ask you to teach a rule. Pick one fraud pattern, write the rule, and say which good customers it will annoy.
| Prompt | The point |
|---|---|
| Write a Radar rule for one fraud pattern | You can be specific without hiding behind a model name |
| How is token theft different from card fraud for an AI company | You know who loses money, and you do not reuse the card story |
| A customer already has a fraud vendor. Why Radar? | You lead with a gap they feel, not with a feature list |
Do not spend the hour on Link or Billing. Stripe loops go wrong when someone preps the company and then meets a Radar interviewer. The annual letter and the 288 Sessions launches are context. The rule is the interview.
Bring a loss story, not a launch recap
Bring a time the loss data killed a story sales was already telling. Name the number, the day you changed your mind, and what you told the field the next morning.
Stripe’s careers page favors people who change their mind when the evidence says they were wrong. Fraud stories harden fast.
A launch timeline with no false positive is a weak sample. Bring the customer you harmed and how you fixed the message.
An individual contributor still owns the rule
You do not need a team. You do need one pattern you can explain to a skeptic, including the customers the rule will catch by mistake.
Years in a title will not replace that. If you are early in the craft, prep the rule and the price, not a leadership speech.
Common questions
- Should you prep card fraud or token theft?
- Both, and the difference is the interview. Card testing is table stakes. Token theft at AI companies is the Sessions 2026 expansion, and it needs its own buyer.
- Will they ask you to invent a price?
- The seat includes a pricing narrative for newer offerings. Have a rule for what you charge for, and a metric that would change your mind.
- How technical does a Radar answer need to be?
- You do not need to write the model. You do need to read the docs well enough to explain one rule and its false positives in plain language.