How we use AI

• Practice and exam grading: open-ended (written) answers in practice, and in the Circle qualifying exam, are scored against a rubric by AI, with notes on what was good and what was missing.
• Mock interviews: an AI interviewer asks the questions and follow-ups, and an AI grader writes your scorecard, including the "mock verdict".
• Story and drill coaching: AI gives feedback on the stories in your story bank and on timed drills.
• Resume reading: if you upload a resume for interviews, AI reads it into sections, so the interviewer can ask about it.
• Question drafting: some practice questions are first drafted with AI. They're held for review and approved by an administrator before students see them. [To confirm for every question.]

Voice interviews also use speech technology (from Deepgram) to turn your speech into text and to read the interviewer's lines aloud. That is speech technology, not the AI that writes your feedback, and it's covered in our Privacy Policy.

• AI can make mistakes, miss a good answer, or sound sure when it isn't. Scores can vary between tries.
• A person doesn't review each AI response, and we don't guarantee human review. [To confirm.]
• Feedback and scores are for practice. They are not career, legal, financial, tax or investment advice, and they don't predict how a real interviewer or employer will react.
• If you think a score or feedback is wrong, email [CONTACT EMAIL].

• Interview scorecards mark the verdict as "Mock verdict (AI estimate)" and say it doesn't predict a real outcome.
• The interviewer is described as an AI interviewer.
• [To confirm: label all AI-graded feedback in practice and exams in the same way.]

We don't use AI to make decisions with legal or similarly significant effects on you, such as hiring or lending. Employers don't see your results unless you share them.

One exception to flag: to join the Circle, you need to reach Interview-Ready, pass a qualifying exam with open-ended answers graded by AI, and score 8/10 or higher on a final mock interview graded by AI. A person doesn't review each score. This only decides access to a community feature, but we've marked it for our lawyer to check whether it counts as an automated decision under privacy laws, and whether we need to offer a human review. [Lawyer to confirm.] If you think you were scored unfairly, email [CONTACT EMAIL].

To produce feedback, we send the relevant text (for example, your answer, interview transcript, resume text or story) to Anthropic, our AI provider. [Lawyer to confirm: under Anthropic's commercial terms it doesn't use content sent through its API to train its models by default.] We don't use your content to train our own AI models. We keep a log of each AI request with the feature, AI model and token counts, but not your answer or the AI's reply. More is in sections 2, 4 and 7 of the Privacy Policy.

To protect against abuse and cost, AI features have daily limits, so they can be unavailable for a while when a limit is reached. Please don't put sensitive personal information you don't want processed in your answers.