Where AI note takers store interview data, and how to find out for yours

Where an AI note taker stores interview data depends on the kind of tool. Meeting platforms keep recordings and transcripts in the organizer's cloud storage: Microsoft Teams in OneDrive or SharePoint, Google Meet in the organizer's Google Drive, and Zoom cloud recordings in the Zoom cloud. Dedicated note-taker services keep audio, transcripts and summaries on their own cloud infrastructure and pass audio and text to subprocessors for transcription and summarization. Local-first apps keep transcripts on the interviewer's computer but may still send audio to a speech-to-text service. In every case, copies then spread into the ATS, email and chat. Vendor details below are as of October 2026, from each vendor's own pages.

Knowing the answer matters for interviews more than for most meetings: these are hiring records with retention duties, candidates can ask for them, and the candidate never agreed to the vendor's terms. This page explains what data a note taker creates, the three storage models, real examples, and how to map your own.

What data does an AI note taker create?

A single interview can leave eight or nine kinds of data behind, and they are not always stored together.

What are the three storage models?

1. Built into the meeting platform

The platform records or transcribes, and the files land in the cloud storage tied to someone's account, usually the organizer's. This is convenient for internal meetings and awkward for interviews: the file sits in a personal drive, shared by default with internal attendees, under that drive's retention rules rather than your hiring records policy.

2. A note-taker service in the vendor's cloud

A separate product, often a bot that joins the call or an app that uploads audio, sends the interview to the vendor's servers. The vendor stores the recording, transcript and summary there, and typically uses other companies (subprocessors) for hosting, speech-to-text and AI summarization. Your data's location is therefore every location on that list, not only the vendor's headquarters.

3. Local-first apps

The app runs on the interviewer's computer and keeps the finished transcript and notes there. Audio usually still has to reach a speech-to-text service during the call, so the useful questions are whether that audio is retained and by whom, and what, if anything, the vendor's server keeps.

Examples: where named tools say interview data goes

Each row is taken from the vendor's own documentation, linked, as of October 2026. Vendors change hosting and subprocessors; check the current page before you rely on a row.

The local-first model is the one we chose for our own product. Interview Signal keeps transcripts, quotes, candidate names, notes and submittals on the recruiter's computer; audio goes to a speech-to-text provider in short chunks during the call and is not retained, and our server keeps only counts and durations. Details are on the security page. Whatever model you use, the questions below are the same.

Why do subprocessors matter so much?

A note taker is rarely one company. The BrightHire list above shows the pattern clearly: one company for infrastructure, several for transcription, several for summarization. Each is a place your candidate's words pass through and, depending on its terms, may be kept for a time. Three questions cover most of the risk:

1. Which subprocessors receive audio or transcript text, as opposed to billing or support data?
2. Do the AI and speech-to-text subprocessors retain inputs, and may they use them to train models?
3. How will you be told when the list changes, and can you object?

The full question set is in the AI note taker security questionnaire, and contract terms that should back the answers are in data processing agreements for recruiting software.

Where do copies go after the call?

The vendor's storage is only the first copy. Within a day, the same interview usually exists in several more places:

• The ATS: a pasted or synced summary, sometimes the full transcript as an attachment.
• Email: automatic summary emails to attendees, forwarded notes, submittals to clients.
• Chat: notes shared in a hiring channel for a quick debrief.
• Calendars: transcripts attached to the meeting event, as Google Meet does.
• Interviewers' downloads: exported Word files and recordings on laptops.
• The client's systems, for agencies: once a submittal is sent, it follows the client's rules.

Each copy needs to be covered by your retention schedule and your deletion cycle. The list of places copies hide, and a schedule you can adapt, are in how long to keep interview notes. For Zoom specifically, the settings that control how long cloud recordings stay are in Zoom recording retention settings for interviews.

How do you find out where your tool stores data?

Check these six places, in this order, and keep copies of what you find with the date.

1. The security or trust page. Hosting provider, region, encryption and certifications are usually stated here.
2. The subprocessor list. The most specific public document: who, for what, where, and when the list was last changed.
3. The data processing agreement. What the vendor commits to contractually, which may differ from the marketing page.
4. The privacy policy. Retention, training and transfer statements, often written for end users rather than business customers.
5. The assurance report. A SOC 2 report describes the system in scope; see SOC 2 for recruiting software.
6. Your own admin settings. Region choices, retention periods, auto-sharing and who can see what in your account.

A data map you can copy

INTERVIEW DATA MAP — [tool] — completed [date] by [name]

DATA ITEM          STORED WHERE (company, country)   KEPT HOW LONG   DELETED HOW
Audio / video      [ ]                                [ ]             [ ]
Transcript         [ ]                                [ ]             [ ]
Summary / notes    [ ]                                [ ]             [ ]
Scores             [ ]                                [ ]             [ ]
Voice profiles     [none / where]                     [ ]             [ ]
Search index       [ ]                                [ ]             [ ]
Backups            [ ]                                [ ]             [ ]

PROCESSED BY (not stored by us):
Speech-to-text     [company, country, retains? trains?]
AI summarization   [company, country, retains? trains?]

COPIES WE CREATE:  ATS [ ]  Email [ ]  Chat [ ]  Calendar [ ]  Downloads [ ]
Covered by our retention schedule?  [yes / gaps: ...]
Sources: [security page URL, subprocessor list date, DPA version]
Next review: [date, or when the vendor changes its subprocessor list]

Why does the location matter for interviews?

• Retention. Interview records have minimum retention periods in many places, and the vendor's defaults become yours unless you change them.
• Access requests. Under laws such as the GDPR and the CCPA, a candidate can ask for their data; you need to know every place it sits. See candidate data access requests.
• International transfers. For UK and EU candidates, storing or processing data in another country brings transfer rules; see recording interviews under UK GDPR.
• Legal holds. If a claim arrives, you must be able to stop deletion everywhere a copy exists, including the vendor's systems.
• Client contracts. Agencies often have clauses on where candidate data may be stored and who may process it.

None of this is legal advice; it is the inventory you need before anyone, including a lawyer, can give you an answer. Whether you may capture the interview at all is a separate question, covered in is it legal to use AI note takers in interviews.