All articles

Transcribing confidential interviews: a practical guide for reporters

A silver Sony microcassette recorder on a dark table, with a microcassette beside it.
Photo: Mk2010 on Wikimedia Commons, CC BY 4.0. Resized.

A microcassette recorder had one security feature that reporters rarely think about: the tape stayed in your pocket. Automatic transcription saves hours, but it can quietly put a copy of the recording somewhere else.

Start with the source, not the software

The Freedom of the Press Foundation’s review of popular transcription tools gives blunt advice: if audio in the wrong hands could put people at risk, avoid transcription services. So the first question is about the person speaking, and how badly a leak could hurt them.

  1. Low

    A public talk, a press conference, your own notes

    Almost any tool will do. Choose for accuracy and convenience.

  2. Medium

    An unpublished interview with a named source, embargoed material

    Keep it local where you can. If you upload, read the retention and training terms first, then delete the file from the service afterwards.

  3. High

    A source who could lose their job, freedom or safety if identified

    Do not upload it. Ask your security team, limit who can reach the file, and consider whether you need a transcript at all.

A starting point for your own policy, not a rule. Your newsroom's security team knows your situation better.

What changes when you stay local

Transcribing on your own device removes one party from the chain. It doesn’t remove every risk, and it’s worth being clear about which ones stay.

RiskCloud serviceOn-device
Provider staff or vendors reach the filePossible, subject to policyNo provider copy
A legal demand reaches the providerPossibleNothing for them to hand over
A breach at the providerYour file may be includedYour file isn’t there
Your phone is seized or stolenDepends on your account securityStill a risk, so use a strong passcode
You export or share the fileYour responsibilityYour responsibility

A workflow for a sensitive interview

  • Prepare before you leave. Download the model you want while you have a connection. After that, you don’t need one.
  • Lock the phone down. Set a strong device passcode and turn on the app lock. Consider airplane mode so you can see that nothing is connected.
  • Tell your source how it works. Say where the recording will live and who can reach it. Check the recording consent rules where you both are.
  • Check quotes against the audio. Automatic transcripts make mistakes, and the mistakes can read smoothly. Tap the word and listen before you quote anyone.
  • Treat exports as a new copy. A file you export leaves the encrypted library, so it needs its own protection and its own place in your retention plan.
  • Decide when to delete. Agree this with your editor, and delete the recording and transcript when that time comes.

If you do use a cloud service

The same review has practical advice for lower-risk work. Use a long, unique password from a password manager, and turn on two-factor authentication if the service offers it. Limit who can reach the data, and delete transcripts from the service once you’ve safely downloaded them.

Your local options

EchoLocal isn’t the only way to stay on-device. The foundation also points to running OpenAI’s Whisper yourself, Google Recorder on Pixel phones and the on-device transcription in iPhone Voice Memos. These differ in languages, export formats and how much setup they need, so try a few on recordings you don’t mind losing.

EchoLocal runs on iPhone and Android, supports 19 languages and keeps a timestamp on every word, so checking a quote takes one tap. It has no accuracy guarantee, and no tool should be your only check. Try your own audio on the free models first.

Sources

Checked on 5 November 2026. This article is general information, not legal, medical or security advice.