Privacy & Security
at Peatch AI
Your interview data deserves the highest level of protection. We've built Peatch AI with enterprise‑grade security from the ground up.
Our Security Philosophy
At Peatch AI, we believe that trust is earned through transparency and robust technical measures. We handle sensitive career data, resumes, interview transcripts, voice recordings, and feedback, and we take that responsibility seriously.
Our security architecture follows the principle of defense in depth: several layers of protection, from access control in the app to encryption at rest and in transit.
Encryption at Rest and in Transit
Our hosting providers encrypt all stored data at rest, and every connection is encrypted in transit. On top of that, Peatch encrypts each candidate’s personality answers and result with AES‑256‑GCM, using a key unique to that candidate. Interview transcripts, recordings, CV text and feedback rely on our hosting providers’ encryption.
AES‑256‑GCM Encryption
Personality answers and results are stored encrypted with AES‑256‑GCM in our database. Everything else Peatch stores is encrypted at rest by our hosting providers.
- Personality answers and results encrypted with a key unique to each candidate
- Transcripts, CV text and feedback kept in a database encrypted at rest
- Recordings, CV files and uploaded documents kept in object storage encrypted at rest
- Every connection encrypted in transit
Per‑Candidate Key Derivation
Using HKDF, we derive a separate key for each candidate’s personality result, so one key never opens another candidate’s result.
- Each candidate’s personality result has its own key
- The master key is kept outside the database
- Keys derived from master secret using cryptographic best practices
- 12‑byte random nonce per encryption operation
Email Pseudonymization
Your email is hashed using SHA‑256 to create a pseudonym that protects your identity in operational systems.
- One‑way hash cannot be reversed
- Enables identification without exposing email
- Analytics and logs use pseudonyms only
- Consistent hashing for reliable identification
Strict Access Control
Role‑based permissions ensure only authorized users can access sensitive data.
- Candidates can only access their own sessions
- Hiring managers limited to invited candidates
- Share tokens provide time‑limited, scoped access
- Cross‑user data access is technically impossible
Technical Details
- Algorithm: AES‑256‑GCM
- Key Derivation: HKDF with SHA‑256
- Nonce: 12‑byte random per operation
- Authentication: Built‑in GCM tag
Strict Access Control
We enforce rigorous access controls to ensure only authorized users can view sensitive data.
Session Ownership Verification
Candidates can only access their own sessions. Hiring managers can only view sessions from candidates they've invited.
Role‑Based Permissions
Every API endpoint verifies user identity and role before returning data. Cross‑user data access is technically impossible. The only exception is a page a candidate publishes on purpose: their Super CV shows just the sections they switched on, to anyone holding the link, and stops working the moment they unpublish it.
Rate Limiting & Abuse Prevention
We implement endpoint‑specific rate limiting to prevent abuse and protect against brute‑force attacks.
| Endpoint | Limit | Purpose |
|---|---|---|
| Password Reset | 3/hour | Prevents email flooding |
| Magic Link Login | 5/hour | Prevents account enumeration |
| Share Token Validation | 10/min | Prevents token guessing |
| Interview Conversation | 60/min | Fair usage enforcement |
| Resume Upload | 10/hour | Prevents storage abuse |
| Account Creation | 5/hour | Prevents spam signups |
Our rate limiter uses a fixed‑window algorithm and returns standard HTTP headers:
X-RateLimit-LimitX-RateLimit-RemainingX-RateLimit-ResetComprehensive Audit Logging
Every access to sensitive data is logged for security monitoring and compliance.
What We Log
- Who accessed the data (user ID and type)
- What was accessed (resource type and ID)
- Which fields were retrieved
- When the access occurred
- Request metadata (anonymized)
What We Don't Log
- The actual content of encrypted fields
- Plaintext email addresses
- Full IP addresses (we keep only a truncated form)
- Passwords or authentication tokens
Privacy‑Preserving Analytics
We use PostHog for product analytics, but with strict privacy controls.
Memory‑only persistence
No analytics data stored in localStorage or cookies
Property denylist
Fields like email, name, phone, resume, password, and API keys are automatically excluded
No cross‑session tracking
Each session is independent, no persistent identifiers
Data Retention Policies
We only keep your data as long as necessary, and we're transparent about our retention periods.
| Data Type | Retention | Action |
|---|---|---|
| Interview and Meeting Recordings | 1 year after the interview or meeting | Permanently deleted |
| Resume Content | 1 year | Anonymized |
| Personality results | Until you delete them or your account | Deleted |
| Personality requests from companies | Until you erase your account, or the company deletes the application, the position or its workspace | Deleted |
| Password Reset Tokens | 1 day | Permanently deleted |
| Session Transcripts | Account lifetime | Deleted on account deletion |
| Unconfirmed early access requests | 7 days | Permanently deleted |
| Confirmed early access requests awaiting a decision | 90 days | Permanently deleted |
| Declined early access requests | 30 days | Permanently deleted |
| Approved requests never activated | 30 days after the link expires | Permanently deleted |
| Questions about your Super CV and their answers | 1 year | Permanently deleted |
GDPR‑Ready Architecture
We've built our platform with GDPR compliance in mind, ready for EU expansion.
Right to Access
Export all your data in a machine‑readable format
Right to Rectification
Update or correct your personal information
Right to Erasure
Request complete deletion of your account and data
Right to Portability
Download your data to take elsewhere
Infrastructure Security
Our infrastructure is designed with security as a first principle.
Deployment
- Platform: Fly.io with isolated containers
- Database: Managed PostgreSQL with automatic backups
- In transit: TLS for web traffic, DTLS-SRTP for live calls over WebRTC
- Secrets: Environment variables stored in secure vault
Startup Validation
- Encryption key length and format verification
- Required secret presence checks
- Database connection check
- Graceful degradation with warnings
Early access requests
When you ask to try Peatch, we store your name, your work email, your company details and your answers about the role you want to fill, together with the language you used and how you reached the form. Until you send your request, your answers stay only in this browser tab and disappear when you close the tab.
- We process this information to take the steps you asked for before a contract: reviewing your request and setting up your workspace (GDPR Art. 6(1)(b)).
- Unconfirmed requests are deleted after 7 days, confirmed requests we have not decided on after 90 days, declined requests after 30 days, and approved requests whose activation link was never used 30 days after that link expires. Once you activate your access, your request is kept with your workspace. The emails we sent about the request are deleted with it. You can ask us to delete your request sooner at any time; we then erase it together with the emails we sent about it.
- Resend delivers our emails. Once you confirm your email, Slack shows our team your first name, your company name, size and country, the first role you want to fill and its level, how many hires you plan, what slows your hiring down the most and the applicant tracking system you use, plus the domain of your email address and whether it is a personal mail provider. Slack never receives your full email address or your last name.
- We never use your request for marketing.
Data from your Google or Microsoft account
If you connect your Google or Microsoft account, Peatch accesses only the following:
- Your email address, to identify the account.
- Your free/busy times in your calendar, to offer candidates your free slots.
- Creating and cancelling the interview events you schedule in Peatch.
- Only if you turn on “Emails to candidates”: sending from your address the emails Peatch writes for your candidates.
Peatch does not read your emails or the content of your other events. This data is used only for these features: it is never sold, used for advertising or used to train AI models.
You can disconnect your account at any time in your settings: Peatch then deletes its access tokens.
Peatch’s use and transfer of information received from Google APIs adhere to the Google API Services User Data Policy, including the Limited Use requirements.
Your personality assessment
The personality assessment is optional. It describes how you prefer to work, and the result is yours: a company sees it only if you choose to share it with that company.
- Software, not AI, scores your answers. Anthropic’s Claude then writes a short description of your working style from those scores, and to write it receives only the scores: never your individual answers, your name or your contact details.
- When you talk to the AI Companion, Claude also receives your profile type to tailor its advice. A company you share your result with can let Peatch’s AI assistant, or an AI app it connects to Peatch, read that result, without your individual answers.
- AI never decides anything about your application. People at the company do, and saying no never counts against you.
- You choose, company by company, who sees your result. A share you choose before taking the assessment takes effect only once your result is saved, and you can stop sharing at any time from your profile.
- A company can ask you to take the assessment or to share your result. The request comes from Peatch, its link works once and for 7 days, and opening it changes nothing until you choose to continue.
- A company that asked sees when it asked. It never learns whether you took the assessment and kept it private or never took it.
- Your answers and your result are encrypted with a key unique to you.
- We keep your result until you delete it from your profile or delete your account. It rests on your consent, which you can withdraw at any time.
Third‑Party Services
We carefully select partners who share our commitment to security.
| Service | Purpose | Data Shared |
|---|---|---|
| Anthropic (Claude) | AI interviews, interview evaluation, meeting summaries and the summary of a personality result | Interview transcript as recorded (encrypted in transit), and, on interviews conducted through OpenAI, the question you were asked and everything you have said in reply since then, sent each time you pause; and the scores of a personality result, to write its summary, your profile type when you talk to the AI Companion, and a result you shared when a company asks Peatch’s AI assistant about it |
| Deepgram | Speech recognition in the voice companion, and speech synthesis for voice previews | Audio of your companion sessions |
| ElevenLabs | Voice agent that conducts the spoken interview, and voice previews | Your live voice during the interview, and the recording and transcript of the interview, which ElevenLabs keeps in the United States after Peatch fetches its copy |
| OpenAI | Voice agent that conducts the spoken interview, when the employer selects it for the position | Your voice during the interview and its transcript, sent outside the European Union to the United States, kept only as long as running the service requires. There is no project keeping this transfer inside the European Union today; that would be a choice for Peatch's own OpenAI account to make, not something a workspace can select. |
| Cartesia | Speech synthesis in the voice companion and for voice previews | Text to be spoken |
| Mistral | Speech synthesis for voice previews and for live voice sessions, when a workspace selects it | Text to be spoken |
| LiveKit | Real‑time audio and video for the voice companion and the meeting room | Live audio and video of those sessions |
| Daily | Video room for meetings with the hiring team, before our own meeting room replaced it in 2026 | Audio, video and transcript of those meetings, deleted from Daily one year after each meeting |
| Recall.ai | Brings Peatch into meetings held on Google Meet, Microsoft Teams or Zoom | Live audio and video of those meetings, the names participants show there, and the recording when they agree, European Union region |
| xAI (Grok) | Speech recognition for meeting recordings and for your practice answers, and speech synthesis for voice previews | Audio of your meetings with the hiring team, and of the answers you record to practice (not kept after transcription) |
| Tigris | Stores interview recordings, CV files and uploaded documents | Audio and video recordings, CV files and uploaded documents, region not yet pinned to the European Union |
| Langfuse | Traces of AI calls, for debugging | Prompts and AI responses, European Union region |
| Resend | Transactional email | Email address only |
| Meta (WhatsApp) | WhatsApp invitations and reminders, when you or the hiring team turn them on | Your mobile number, the employer's name, the position and your interview link |
| PostHog | Product analytics | Anonymized usage data |
| Slack | Internal team notifications | For confirmed early access requests: first name, company name, size and country, first role and its level, hires planned, main hiring slowdown, applicant tracking system, email domain and whether it is a personal mail provider. Never the full email address or the last name. |
| Calendar and, if you turn it on, candidate emails from your Gmail address | Interview events and the emails Peatch sends on your behalf | |
| Microsoft | Calendar, Teams links and, if you turn it on, candidate emails from your Outlook address | Interview events with their Teams links, and the emails Peatch sends on your behalf |
| GitHub | Finds and reads public GitHub profiles of candidates, when a hiring team turns public profiles on | The candidate’s email address in search queries, and public usernames |
| Gravatar (Automattic) | Reads the public Gravatar profile linked to a candidate’s email address | A SHA-256 hash of the candidate’s email address |
Anthropic, ElevenLabs and OpenAI, the three providers that read or conduct the interview itself, process that data in the United States. Anthropic also processes there the scores of a personality result, your profile type when you talk to the AI Companion, and a result you shared when a company asks Peatch’s AI assistant about it.
Questions?
If you have questions about our privacy practices or security measures, we're here to help.
Privacy Questions
[email protected]Security Issues
[email protected]Last updated: November 2026