Anonymize interview transcripts before AI touches them.
AI is genuinely useful for coding, summarizing, and finding themes in qualitative data. But transcripts are full of names, places, and small identifying details — and the rules you work under exist precisely because of that. Data Alias de-identifies qualitative data for AI analysis inside your browser, so the transcript itself never leaves your machine.
The obligation is real, and it is written down
University IRBs have started addressing AI directly: Columbia Teachers College and Georgia State, among others, publish guidance requiring that human-subjects data be de-identified before it is processed by AI tools. If your protocol promised participants confidentiality, pasting a raw transcript into a chatbot is a protocol question, not a convenience question.
European researchers carry a parallel obligation: under the GDPR, pseudonymised data is still personal data — data protection authorities are explicit that re-identification risk remains, which is why pseudonymisation is treated as a risk-reduction measure, not an exit from the rules. Handling the mapping locally, under your control, is the conservative reading of that guidance.
The strange deal cloud de-identification asks you to take
Established de-identification services charge per transcript — commonly in the $12–15 range — and their first step is asking you to upload the very data you are trying to protect. For a 30-interview study that is real money, and every upload is a new custody question for your ethics documentation.
Data Alias inverts the model: the de-identification runs in your browser, transcripts and alias mappings stay on your machine, and you can verify the no-upload claim yourself with the network tab or airplane mode. Flat pricing, not per-transcript.
A workflow built for study data
Drop in DOCX or PDF transcripts
Word documents and text-layer PDFs are extracted directly in your browser — no upload, no conversion service. Each transcript becomes reviewable text with detected emails, phone numbers, and other contact details flagged in place.
One participant, one alias — across every file
The encrypted alias vault keeps a single project mapping, so the same participant is PERSON_004 in every interview file you process. Cross-case analysis, quote attribution, and member checking still work on the de-identified set. The vault is encrypted with your passphrase and lives only in your browser's storage.
A report for the ethics board
Generate an on-device anonymization report describing what was detected and what was transformed — a concrete artifact to attach to your data-management documentation. It documents your process; it does not certify compliance, and we won't pretend otherwise.
A reversible mapping that stays with you
Export the alias mapping and keep it under your study's key file, exactly as re-identification linkage sheets are traditionally handled. When the AI's summary talks about PERSON_004, de-alias it locally — the real names never entered the AI conversation.
What this tool is not
- • Detection can miss things. Interview speech is messy — nicknames, indirect identifiers, places described rather than named. No automated detector catches everything, and we won't claim ours does.
- • The review step is part of the method, not a formality. You read the de-identified transcript before it goes anywhere, the same way you would proofread a redaction.
- • This is not a compliance certification. The report documents what you did; your IRB, ethics board, or DPO decides whether the protocol is satisfied.
Questions researchers ask
- Can I keep the same participant alias across every transcript in a study?
- Yes. The alias vault keeps one mapping per project, encrypted with a passphrase you choose and stored only in your browser, so PERSON_004 means the same participant in interview 1 and interview 12. The mapping is never uploaded.
- What can I show my IRB or ethics board?
- An on-device anonymization report summarizing what was detected and how it was transformed. It documents your de-identification step; it is not a certification, and your board remains the judge of whether the protocol is satisfied.
- Is the de-identification reversible for my own records?
- Yes, locally. You can export the alias mapping and keep it with your study records; the restore page turns a safe copy plus its mapping back into the original — all in your browser, so the linkage file never leaves your machine.
DOCX and PDF extraction are Pro features, free during the beta. See also the Word transcript guide.