Security and data handling
This page says where your data sits, who can reach it and how it is deleted. Each line describes what the product does today. Gaugely is a small, founder-run company, so the last section lists what we do not have yet.
Hosting and where data sits
- The web application runs on Hostinger in Singapore. The database is Neon Postgres on AWS in the United States (US East). The interviewer agent runs on LiveKit Cloud in US East. Recordings and uploaded files are in a private Cloudflare R2 bucket.
- Gaugely is run from Lahore, Pakistan. There is no EU-only or UK-only data option today.
Encryption
- In transit: the site is served over HTTPS only, with a two-year HSTS header that covers subdomains. Neon states that it requires TLS on every database connection. Live audio and video travel over WebRTC, which encrypts media in transit.
- At rest: Neon states that customer data is encrypted at rest with AES-256. Cloudflare states that every object in R2 is encrypted at rest with AES-256. We rely on those providers for encryption at rest and do not add our own layer.
Access control and roles
- A workspace has four roles: owner, admin, recruiter, viewer. A viewer can read and cannot change anything. Only an owner and admin can manage the team, billing, API keys and the public careers page.
- The role is read fresh from the database on every action that writes, sends email or spends interview minutes, so a removed or demoted member loses access on their next action.
- Sign-in needs a verified email address. Seats are limited by plan.
- Recordings are never public. They are streamed only to a signed-in member, and only for files under that member’s own workspace.
- The internal admin area is limited to an allow-list of staff email addresses, and every change made there is written to an audit trail.
Sub-processors
Every outside service that can receive personal data, what it is for, and where it runs.
| Service | What it does and what it receives | Region | When |
|---|---|---|---|
| Hostinger | Hosts the web application. Every request to the site passes through it. No database or recordings are stored there. | Singapore | Always |
| Neon | Managed Postgres database. Accounts, jobs, candidate names and emails, transcripts, scores, consent and event logs. | United States (AWS US East) | Always |
| LiveKit Cloud | Carries the live audio and video of an interview, runs the interviewer agent and writes the recording. Live audio, and video or screen share when the job uses them. | Agent runs in US East; media is carried on LiveKit’s network | During an interview or a built-in video call |
| Cloudflare R2 | Stores interview recordings and uploaded files in a private bucket. Recordings, uploaded CVs and recruiter-uploaded recordings. | Not pinned by us | When a recording or file is stored |
| Deepgram | Turns speech into text and text into the interviewer’s voice. Interview audio and the interviewer’s spoken lines. | Not pinned by us | During an interview; also for recording uploads |
| Groq | Language model for the interviewer, scoring, CV ranking and writing tools. Transcripts, CV text, job descriptions. | Not pinned by us | Always for the live interviewer; first choice for paid interactive work |
| Cloudflare Workers AI | Language model for free-plan and bulk work, and a fallback. Transcripts, CV text, job descriptions. | Not pinned by us | Only when configured; free-plan and bulk requests, or when the first choice is unavailable |
| OpenRouter | Fallback route to the same language model on another host. Transcripts, CV text, job descriptions. | Not pinned by us | Only when configured and an earlier choice fails |
| Resend | Sends email: invites, reminders, results, feedback, privacy notices. Names, email addresses and the email text. | Not pinned by us | Always |
| Polar | Merchant of Record for billing and tax. Recruiter billing details. No candidate data. | Not pinned by us | Paid plans only |
| Didit | Government ID and liveness check before an interview. The candidate’s ID document and selfie go to Didit. We receive the result. | Not pinned by us | Only when a job requires identity verification |
| PostHog | Anonymous page and sign-up measurement on public pages. Page and conversion events. No candidate data, transcripts or form values. | United States | Only after a visitor chooses Allow |
| Google Analytics | Visit measurement on public marketing pages. Visit data. Never runs on sign-in, interview, meeting or app pages. | Not pinned by us | Public pages; in the EEA, UK and Switzerland, analytics cookies only after Allow |
"Not pinned by us" means we have not chosen a region for that service, so its own default applies. The code also supports services that stay off unless they are configured: Sentry for error reports, Cloudflare Turnstile for the sign-up bot check, a code-running service for coding questions, the Better Auth hosted dashboard, and alternative AI models from Anthropic, Google and Mistral. Email support@gaugely.io and we will tell you which are on today.
AI services and your data
Gaugely does not train or fine-tune any AI model on your interviews, CVs or scores. The services below process them to return a result. This is the data-use setting our code sends to each one today.
- Deepgram
- Every request opts the audio out of Deepgram’s model improvement programme (the mip_opt_out setting, developers.deepgram.com, read 2026-11-19). Deepgram’s terms still govern how long it holds data to run the service.
- Groq
- We send no data-use setting. Its standard API terms decide how long it keeps data and whether it may use it.
- Cloudflare Workers AI
- No setting to send. We rely on Cloudflare’s published policy, as we read it on 2026-09-23, that it does not train on or retain customer content.
- OpenRouter
- Every request carries a setting that excludes hosts which collect inputs.
Retention and deletion
- Each workspace has a retention window, 24 months by default, counted from a candidate’s last activity.
- Automatic erasure at the end of that window is built and is switched off by default. It stays off until deletion from file storage has been verified on the live bucket. Until then a scheduled job only reports who would be erased.
- Erasure on request works today. A candidate can ask from the link in their privacy email or by writing to support@gaugely.io. The employer, who is the controller, acts on it in Settings and has 30 days. Erasure removes the person’s details and stored files and leaves an anonymous marker so the record of the request survives.
- A candidate who joined the talent pool stays in it only while opted in and can withdraw at any time.
Report a vulnerability
- Email support@gaugely.io with "Security" in the subject. Say what you found and how to reproduce it. A person reads that inbox. There is no bug bounty, and we will not take action against good-faith research that avoids other people’s data.
What we do not have yet
- No SOC 2 or ISO 27001 report.
- No penetration test report from an outside firm.
- No independent bias audit of the AI scorer, and no measured agreement with human raters yet.
- No single sign-on (SSO) and no enforced two-factor sign-in.
- No uptime commitment (SLA) and no public status page. The web application sleeps when idle, and the interviewer agent can take 10 to 20 seconds to start.
- No signed-DPA download. We send one on request.
If your buying process needs one of these, tell us before you start so nobody's time is wasted.