Is ElevenLabs safe to use?
AI audio company. Text-to-speech, voice cloning, speech-to-text, dubbing, conversational agents, and related creative tools — including the voices many apps, and this site’s cast, use for character audio.
Last reviewed: 24 August 2026
Does ElevenLabs train on what you type?
Yes, unless you switch it off — and you can.
Can you turn ElevenLabs's training off?
Stop new data going into model training: open ElevenLabs, click your profile icon, go to Terms and privacy → Data use, turn off Improve the models for everyone, then click Update your choice. After that, new data you submit should not be used to train. Older Typeform opt-outs should already show in that dialog. Treat your voice like a house key, not a throwaway demo. Only upload recordings of voices you have the right to use; do not upload anyone under 18. If you only needed a one-off test, delete the voice or clone and generations you no longer need from the product, and use account/data-deletion paths if you are done with the service entirely. If you handle sensitive scripts — health, banking, client confidential — consumer history defaults are not zero retention. Enterprise customers can ask about no-train-by-default contracts and Zero Retention Mode on eligible API and agent products. ZRM does not cover Instant or Professional Voice Cloning sample storage. Voice Library: only publish a voice model there if you are comfortable with other users generating speech in that voice.
Is what you put into ElevenLabs private?
ElevenLabs processes what you type, upload, or record to run the product: text prompts, generated audio, and Voice Data, meaning information about your voice derived from recordings you provide. If you build a voice clone, those samples are analyzed so the service can generate speech that sounds like that voice. Voice Data may count as biometric information under some laws; ElevenLabs says it does not use Voice Data to infer traits about you as a person. On consumer and other non-Enterprise accounts, ElevenLabs may use certain data you provide to improve its audio models for everyone. It says it does not use your data to clone your voice without consent or to build advertising profiles of you. Enterprise customers are different: by default, ElevenLabs says it does not train on Enterprise customer data except as needed to provide the service. Anyone can turn off training use for new data through the account Data use control. Opting out does not automatically erase past training use of data you already sent. Separate from model training: generation history and voice models are retained under ElevenLabs’ retention rules so the product works, and for security or troubleshooting. Voice-related generated data is described as kept no longer than about three years after your last interaction, except where law requires longer. Instant and Professional Voice Cloning samples are not covered by Enterprise Zero Retention Mode. If you publish a voice into the Voice Library, other ElevenLabs users can generate audio with it and a preview sample is shown.
So, is ElevenLabs safe to use?
Mostly, with care. We currently rate ElevenLabs "Use With Caution" (privacy risk: Medium). Worth reading the points above before you share anything personal.
The Skeptic's take
“I kept thinking, ‘it’s just a voice sample for a joke video.’ It is also a recording of me sitting on someone else’s servers. I turned off Improve the models for everyone before uploading anything I would not want in a shared model — and I still would not clone my kid, my boss, or a stranger for fun.”
— The Skeptic
The official documents behind this review
We read these so you don’t have to. Check them yourself any time.
Wondering how we decide these ratings? Read how we review AI privacy.
Knowing what’s safe is half of it.
The other half is knowing what to do with these tools. We teach that part — tutorials, prompts, plain English, no judgment. For less than a cup of coffee a month.
Show Me How This Works →Start free. Cancel anytime. No judgment, ever.