Voice assistants like Alexa raise obvious privacy concerns, and have since the moment after they were introduced and we got over the ‘wow cool’ phase. When it was recently revealed that humans were auditing some recordings—on Alexa, Google, and Siri—the concerns grew even louder.
Despite the apprehensions, voice-activated devices and capabilities like these are becoming more pervasive, so we thought it was a good time to take a broad view of just what risks you face if you allow these devices to sit in your home.
The Privacy Risks of Voice Assistants
Voice assistants present a tough place to try and strike a balance between privacy and convenience. They’re handy things to have around, but they pose a fundamental privacy risk. You’re placing an always-on microphone in direct proximity to the most private areas and parts of your life. And once it’s there, you invariably forget about it But it’s listening 24/7/365.
There are three fundamental risks:
- The right recordings are made, but they get used in ways that aren’t intended.
- They record things they’re not supposed to.
- The device is hacked and used to listen in ways the manufacturer didn’t intend.
Let’s consider each of those:
1. People Hear Your Recordings
Despite the voice assistants’ pleasing human voices, we all assume we’re talking to a computer. We further assume that no people will listen to the recordings these devices make. But this isn’t entirely reasonable. You shouldn’t be surprised when the companies building and refining these devices want to improve them, and might want to use data from current users to make a better product.
Say you’re looking for a recipe, directions, or want to listen to a specific song. If the voice assistant doesn’t capture the audio correctly, it has failed.
The only way to improve the performance is to have humans flag false positives and manually note when and how a voice assistant was faulty. Otherwise, the assistant’s performance would never improve.
And in fact, until recently that’s exactly what happened. Amazon, Google, and Apple all had people who were paid to listen to recordings from customers, compare them to computer interpretations of what was said, and use that to help engineers make the voice recognition work better (especially for someone with an accent or dialect). Amazon hasn’t released its figures, but says only a small sample of recordings are listened to; Google says it only listens to 0.2% of recordings; Apple listens to less than 1% of recordings.
Since the public revelation of these practices, each company has changed the way it handles the opt-out process for this kind of review.
The other risk, much smaller, is that someone hacks into the servers at Amazon, Google, or Apple and gets to listen to your recordings, or share them publicly. To guard against that you have to largely trust these vendors to have solid security, but each device also has settings and options to delete your old recordings.
(To learn more about the human review process Amazon and other companies conducted as well as what their current process is, check out this article.)
2. You Get Recorded When You Didn’t Ask To Be Recorded?
Anyone who owns one knows that voice assistants accidentally turn because they incorrectly ‘hear’ their prompt to turn on (“Alexa!”, “Hey Siri!”). Public evidence shows this has resulted in private and compromising conversations to be recorded.
This doesn’t happen a lot, and the chance of it happening when you’re saying something particularly private or personal is even smaller, although that obviously depends on how often you have sensitive conversations.
But there is not much you can do about this risk if you’re going to own these devices. One option is to temporarily disable them when you want to make sure you’re not being recorded. You can unplug some devices or turn off the microphone so it can’t record anything you’re saying.
3. Your Device Gets Hacked, and Starts Recording Everything
What if someone could turn your Voice Assistant into a true open mic, allowing them to listen to anything, anytime, or to record everything and send them a copy? That’s clearly the nightmare scenario—a risk that only exists because you put a 24/7 microphone connected to the internet into your living room… or bedroom.
There is no evidence this has happened, but it’s definitely possible—even probable. Certainly the developers know that a ton of these devices will be in the garbage the day after this is publicly revealed, so they’re highly motivated to make sure it doesn’t happen, but there is no 100% security.
If this risk worries you, don’t keep one around, or disable or better yet unplug it when you don’t plan to use it or when you’re having sensitive conversations.
In order to better understand the impact of an issue and realize it’s implication, ask an important question: ‘What’s the risk?’ Then ask the same thing whenever privacy issues pop up.
Whether someone should own an Alexa, use Siri, or Google Assistant is a personal decision and it should be done with all the relevant information considered. There’s no right answer and even the question of whether these devices are good for privacy is a personal matter. Understanding that privacy can sometimes come at the cost of convenience is an important framework to use whenever trying to figure out how practical your privacy can be.