Everything you say to Alexa now becomes training material for AI, and no, you cannot opt out

Isn’t it convenient to live with a voice assistant? You can go about your day and to perform tasks hands-free — ask to set a timer or remind you of something, add items to your shopping list, or get quick information.

And they’re getting smarter and smarter. Alexa has even learned to mimic the voices of deceased relatives — a controversial feature, both ethically and in terms of privacy. With the rise of generative AI, voice assistants are becoming even more advanced, taking a more personalized approach to every user and raising new questions on the privacy front.

Previously these mounting concerns were alleviated by the fact that the voice recordings could be processed locally on your device. You had the choice to prevent your voice data from being sent to Amazon’s servers. But as of March 28, 2025, Amazon has removed the "Do Not Send Voice Recordings" option for several Echo devices, including the 4th-generation Echo Dot, Echo Show 10, and Echo Show 15. This change means that all voice commands — and effectively everything your Echo hears — will now be transmitted to and processed in Amazon’s cloud.

According to Amazon, this change is necessary to power new generative AI features in Alexa. The voice data will help train Alexa+, Amazon’s next-generation voice assistant, launched in early 2025.

Privacy concerns

While Amazon claims that voice recordings are encrypted and deleted after processing, the update has sparked renewed privacy concerns among users.
We’ve already addressed many of these issues before, but as the situation has evolved, it’s worth revisiting some of the concerns here.

Issue 1: Data leaks

Previously, voice data was mostly stored on the user's device, which meant any potential breach was limited to your personal hardware. Now the local processing will be impossible, and all data will be sent to the Amazon servers and stored in the cloud.

This increases the risk of exposure dramatically, turning what was once a personal device-level concern into a centralized cloud-based vulnerability. A single breach could now compromise the data of millions.

Issue 2: Real people might be listening

How exactly will Alexa’s new AI-powered processing and training work? Will it be fully automated, or will humans be involved? This isn’t just paranoia — there have been precedents. Back in 2019, it was revealed that thousands of Amazon employees had access to and were listening to users’ Alexa recordings to "improve the service."

“We only annotate an extremely small sample of Alexa voice recordings in order [to] improve the customer experience. For example, this information helps us train our speech recognition and natural language understanding systems, so Alexa can better understand your requests, and ensure the service works well for everyone.”

–Amazon spokesman

With Alexa+ now training on even more data to support generative AI capabilities, it's unclear whether human review has been scaled back or scaled up.

Issue 3: Data deletion

What goes on the cloud stays on the cloud? Amazon claims that you're in control — and if you remain opted out of saving voice recordings (it is disabled by default), they claim your data will be deleted right after Alexa finishes processing them. This is a quote from the email that Amazon sent to impacted Echo users:

If you do not take action, your Alexa Settings will automatically be updated to 'Don't save recordings.' This means that, starting on March 28th, your voice recordings will be sent to and processed in the cloud, and they will be deleted after Alexa processes your requests. Any previously saved voice recordings will also be deleted.

But here’s the catch — when does that processing end? What exactly does “processing for algorithms training” entail? And how can you check that all your data was actually deleted after the said “processing”?

Conclusion

At the end of the day, companies don’t just collect your data — they thrive on it. As generative AI becomes the next big battleground, your voice, habits, and daily interactions are more valuable than ever. What used to be private commands whispered to a device in your kitchen are now training data for powerful machine learning models.

Amazon isn’t alone in this shift — nearly every tech giant is racing to gather more user data for the sake of “improving AI”. But what is the cost for us, the users? The cloud may be invisible, but the consequences of what you share with it are very real.

Surely, the only way to protect your data 100% is to not use any digital services at all — but that’s clearly not what we’re suggesting. Ultimately, the decision to trust a company with your data is yours alone. Our goal is simply to shed light on the risks involved so that your choice is informed and thoughtful. And in times when convenience often comes at the expense of privacy, staying informed is as important as ever.