Trust
Subprocessors
These are the outside services that handle interview, candidate or account data for us. For each one: what it does, the data it handles, where it runs, and whether it is always on or only a fallback. Where a location is not fixed by our own settings, we write "Ask us" rather than guess.
Account holders are emailed before this list changes. The privacy policy names the same providers, and the security page explains how each kind of data is handled.
Last updated: November 4, 2026
| Provider | What it does | Data it handles | Where | Use |
|---|---|---|---|---|
| AI models | ||||
| Anthropic (Claude) | Writes interview reports and the AI interviewer's replies. Reads candidates' messages in the hiring assistant chat and answers questions about the job. Answers account holders in the help chat. | Interview transcripts, the job description, the interview plan and resume details. In the hiring assistant chat, the candidate's typed message and the job's facts, without the candidate's name, contact details or resume. In the help chat, the account holder's question, with no candidate data. | Ask us | Always on |
| Google (Gemini) | Reads resumes into fields, suggests live follow-up questions and writes interview questions. Writes the report if Claude cannot, and is the last backup for the AI interviewer's replies. | Resume text, the job description, the interview plan and transcript text. | Ask us | Always on Resume reading, follow-up suggestions and question writing run on it by default. Report writing and the AI interviewer's replies use it only when Claude fails. |
| OpenRouter | A gateway to other AI models. It carries interview-question requests when a firm picks a model other than our default, and the AI interviewer's reply when Anthropic has not started to answer within 3 seconds and our backup route is switched on. | For question writing, the job and interview-plan details, and for personalised questions an extract of the resume. For the backup route, the transcript text of the candidate's speech, the job description, the interview plan and a resume summary. On every request we ask for zero data retention and refuse hosts that store prompts; a question-writing model with no such host is not used, and our default model writes the questions instead. | Ask us | Optional Used only when a firm picks another question-writing model, or when the backup route is switched on and Anthropic is slow. |
| Amazon Web Services (Amazon Bedrock) | Hosts the Claude model for the AI interviewer's backup route, reached through OpenRouter. | Transcript text of the candidate's speech, the job description, the interview plan and a resume summary. | United States (Bedrock us-east-1) | Fallback only Only when the backup route is switched on and Anthropic has not started to answer within 3 seconds. |
| Google Cloud (Vertex AI) | Hosts the Claude model for the AI interviewer's backup route, reached through OpenRouter. | Transcript text of the candidate's speech, the job description, the interview plan and a resume summary. | United States (Vertex AI us multi-region) | Fallback only Only when the backup route is switched on and Anthropic has not started to answer within 3 seconds. |
| OpenAI | Models that may write interview questions when a firm picks one, reached through OpenRouter with zero data retention. | The job and interview-plan details used to write questions, and for personalised questions an extract of the resume. | Ask us | Optional Only when a firm picks one of its models for question writing. |
| xAI | Models that may write interview questions when a firm picks one, reached through OpenRouter with zero data retention. | The job and interview-plan details used to write questions, and for personalised questions an extract of the resume. | Ask us | Optional Only when a firm picks one of its models for question writing. |
| Speech and voice | ||||
| Deepgram | Turns interview speech into text, as live captions, for the AI interviewer to hear, and again from the finished recording. Gives the AI interviewer its voice. | Interview audio, and the text of the interviewer's lines that it speaks. We switch on Deepgram's model-improvement opt-out for every request. | Ask us | Always on |
| ElevenLabs | A backup voice for the AI interviewer. | The text of the interviewer's lines that it speaks. | Ask us | Fallback only Only if we switch the voice provider from Deepgram to it. |
| Video calls | ||||
| LiveKit | Runs the live video calls and saves each recording to our storage. | Live audio and video of the interview room. | Ask us | Always on |
| Resend | Sends email about accounts and interviews, including sign-in codes, and delivers messages from the contact form to us. | Recipient names and email addresses, interview times and links, sign-in codes, and the name, email, company and message from the contact form. | Ask us | Always on |
| Google Workspace | Hosts our own mailbox, where messages from the contact form and email sent to our support and privacy addresses arrive and are kept. | The name, email, company and message from the contact form, and the content of email sent to us, including any candidate data an account holder puts in a support request. | Ask us | Always on |
| Error monitoring | ||||
| Sentry | Records errors and slow requests from the website, the API, the AI engine and the voice interviewer so we can fix them. | Error details and logs, which can sometimes include personal data. The API leaves out the body of sign-in requests, and page session recording is off. | United States | Always on |
| Hosting and storage | ||||
| Vercel | Hosts the website and the web app. | Web requests, including IP address and browser type, and the pages it serves, such as a shared report. | Ask us | Always on |
| Railway | Runs our API, the AI engine and the voice interviewer. | All application data while it is being processed, and service logs. Recordings and resumes are not stored on Railway. | San Francisco, United States (region sfo) | Always on |
| Neon | Our database. | Accounts, candidates, resume details, transcripts, scores, reports, hiring assistant chats and the audit log. | US East, Northern Virginia (AWS us-east-1) | Always on |
| Upstash | Short-term data storage (a cache) and the queue for background jobs. | Sign-in codes, the details of queued jobs such as a reminder email's recipient address and join link, rate-limit counters and locks. | Ask us | Always on |
| Cloudflare R2 | Stores interview recordings, resumes and firm logos, and the nightly backups: an encrypted database dump and a second copy of the recordings, each in its own bucket. | Interview recordings, resumes, firm logos and the backups. | Ask us | Always on |
| GitHub | Runs the nightly backup jobs. The job dumps the database, encrypts the dump and uploads it to Cloudflare R2, and copies new recordings to the backup bucket. | A copy of the database while the job runs, which the job encrypts before it uploads, and each new recording as it is copied. | Ask us | Always on |
This list covers the services that handle interview, candidate or account data. Services that only check that our site is up do not see that data and are not listed.
To ask where a provider runs, to see the Data Processing Agreement, or to object to a provider, email admin@hireanex.com.