Smart Speakers, Kids, and Data: What Parents Don't Know

A seven-year-old in suburban Atlanta asks Alexa what the weather is going to be. Then she asks Alexa to play a song. Then she and her friend spend twenty minutes asking Alexa random questions and laughing at the answers. Her parents, in the next room, do not think about what is happening. They think about it less than they would if their daughter were on a phone or tablet — the smart speaker is somehow background technology, ambient, furniture almost. What they don’t know is that Amazon retains voice recordings for every interaction, that those recordings may include their daughter’s voice saying identifiable information, that Amazon’s employees and contractors have access to those recordings for quality review, and that the FTC found in 2023 that Amazon had been retaining children’s voice data in violation of its own deletion policies for years. This is the smart home privacy gap: devices that feel passive are actively collecting, and the data they collect from children sits in a legal and regulatory gray zone that parents rarely understand.

Key Takeaways

• Smart speakers are present in approximately 35-40% of U.S. homes with children as of 2025, according to Edison Research and NPR data.
• The 2023 FTC enforcement action against Amazon required $25 million in penalties specifically for Alexa’s retention of children’s voice recordings and location data in violation of COPPA and Amazon’s own privacy policies.
• Smart speakers are not classified as “directed to children” under COPPA, even when they have child-specific features, unless the manufacturer specifically designates the product for children — which most don’t.
• Voice recordings of children have different privacy implications than typed search data: voice carries emotional content, health information, and identifying characteristics that text does not.
• Accidental purchases by children, access to age-inappropriate content, and voice-based behavioral profiling are documented problems with current smart home configurations.
• A set of specific settings changes — available on every major smart speaker platform — can meaningfully reduce data collection without eliminating device functionality.

The 2023 FTC Action Against Amazon: What It Actually Found

The Federal Trade Commission’s 2023 enforcement action against Amazon is the most significant regulatory event in smart home children’s privacy to date, and its findings deserve careful attention because they are specific and documented rather than hypothetical.

The FTC’s complaint, filed jointly with the Department of Justice, alleged three categories of violations related to Alexa and children. First, Amazon retained voice recordings of children for years after parents had requested deletion, and in many cases indefinitely — the recordings were not deleted as parents were told they would be when they used the in-app deletion feature. Second, Amazon used children’s voice data to train its Alexa voice recognition models, even when the recordings came from households that had not consented to training data use. Third, Amazon retained children’s geolocation data collected through Alexa-connected services in violation of both COPPA and its own stated privacy policy.

The $25 million civil penalty was the largest ever levied under COPPA at that time. Amazon did not admit to the allegations but settled and agreed to delete children’s voice recordings, stop using them for training, and implement a proper deletion process going forward.

What the FTC action revealed about the technical architecture of smart speakers is important: Alexa does not process voice locally on the device. When a child speaks to an Alexa device, the audio is transmitted to Amazon’s servers, processed, and stored — including the audio before the wake word (which devices buffer continuously to detect activation). The resulting data is voice recordings, which carry information that text records do not: tone, emotional state, background audio (other family members’ conversations, television content, ambient information about the household), and the child’s voice characteristics themselves (which can be used for voice identification).

The FTC’s investigation found that Amazon’s internal data retention practices did not match what the company told consumers. This is the core finding: the privacy control you are told you have is not the privacy control you actually have. This pattern — a discrepancy between stated deletion and actual retention — has been observed in app privacy research as well, but it is particularly significant for smart speakers because the data category is voice and the population affected includes young children who have no awareness they are being recorded.

COPPA’s Gap: Why Smart Speakers Fall Through

COPPA requires verifiable parental consent before collecting personal information from children under 13 on websites and online services “directed to children.” Smart speakers fall through this framework for a reason that is structural, not incidental.

Amazon Echo, Google Nest, and Apple HomePod are marketed as general-audience household devices. They are not classified as directed to children under COPPA’s criteria, despite the fact that they have child-focused features (Amazon Kids mode, Google Nest’s Family Link integration), are regularly used by children, and collect children’s voice data in the normal course of operation. Because they’re general-audience devices, COPPA’s consent requirements don’t apply — even though children in the household interact with them.

The child-specific modes (Amazon Kids mode, available through Amazon Parent Dashboard) do require parental consent to activate. But these modes are optional, not default. A parent who does not know about Amazon Kids mode — or who knows about it but doesn’t use it because the device is shared by the whole household — has no COPPA protections for their child’s data. The data is being collected; the law simply doesn’t apply.

Research by Pridmore and Mirmalek (2021), published in the journal Surveillance & Society, analyzed smart speaker data practices across Amazon, Google, and Apple and found that all three companies’ default configurations collected voice data from child users in ways that would be prohibited for app developers under COPPA. The researchers characterized this as a category error in the regulatory framework: COPPA was written for apps and websites, and its “directed to children” test does not map well onto ambient household devices that children encounter as household objects.

The FTC’s own 2023 staff report on COPPA’s effectiveness acknowledged this gap explicitly, noting that “ambient computing devices present novel challenges for COPPA’s existing directed-to-children framework” and flagging the gap as a priority for potential rule revision.

What Voice Data From Children Actually Contains

Understanding why children’s smart speaker data is sensitive requires understanding what voice recordings contain — which is considerably more than the surface content of what is said.

A voice recording contains the spoken content (what was requested or asked), the child’s voice characteristics (which can be used for voice identification and age estimation), emotional tone (anxiety, excitement, distress are detectable in voice patterns at accuracy rates well above chance), and background audio from the environment. Research in computational paralinguistics — specifically work by Schuller et al. published in IEEE journals — has demonstrated that emotion recognition from voice is reliably achievable with current technology, and that children’s voices can be age-estimated within 2-3 years from a single sentence.

The behavioral profiling implications are significant. A smart speaker that records a child asking about a health symptom, expressing anxiety about school, or repeatedly asking questions about a topic has data that is qualitatively more sensitive than app usage data. A child who asks Alexa “what is depression” at 11 PM three nights in a row has generated data with clinical significance. This data is not protected differently from other Alexa data, and there is no HIPAA-equivalent protection for smart speaker health queries.

Accidental purchases are a documented and widespread problem that intersects with children’s smart speaker use. According to Consumer Reports and FTC complaint data, children making unintentional purchases through voice-enabled devices have generated millions of dollars in disputed transactions. Amazon’s Echo allows purchases by voice using the account holder’s payment method, and while PIN protection is available, it is not enabled by default. A child who has learned to use Alexa for music and weather information can, without understanding what they are doing, order physical products that arrive at the door.

Smart Speaker Data Category	What Is Collected	Default Retention	Child-Specific Risk	COPPA Coverage
Voice recordings	Full audio including wake word buffer	Indefinite unless deleted	Voice ID, emotional tone, health content	Not covered on general-audience devices
Voice transcripts	Text conversion of every query	Indefinite unless deleted	Search history, behavioral patterns	Not covered on general-audience devices
Device interaction patterns	When, how often, what types of queries	Indefinite	Behavioral profiling	Not covered
Location data (Alexa skills)	Location from connected skills	Varies by skill	Home address, school address	Varies; FTC found Amazon violations
Child voice profiles	If enabled, voice models built per person	Until deleted	Voice identification	Required consent if feature is child-specific
Skill/app usage	Third-party Alexa skills used	Varies by developer	Depends on skill type	Not covered for general-audience skills

What Settings Actually Protect Children

The settings available on major smart speaker platforms vary significantly in what they protect, and most default configurations are set to maximum data collection. The following is specific to the major platforms as of early 2026.

Amazon Echo / Alexa: Enable Amazon Kids mode for children’s interactions — this provides a child-oriented interaction mode, turns off purchasing, filters content, and applies COPPA-compliant data handling to the child profile. To reduce retention of voice recordings for all users, go to Alexa Privacy settings (accessible via the Alexa app) → Review Voice History → Manage Your Alexa Data → set automatic deletion to “3 months” or delete manually. Note: as the FTC found, Amazon’s deletion may not be immediate or complete; this is a mitigation, not a guarantee. Disable voice purchasing or require a PIN: Settings → Voice Purchasing → Require confirmation code.

Google Nest: Family Link integration allows parent supervision of child accounts. Audio activity auto-delete can be set to 3 or 18 months at myaccount.google.com → Data & Privacy → Web & App Activity. “Guest Mode” prevents Voice Match recognition and reduces personalization. Disable purchases: Settings → Payments → Off.

Apple HomePod: Apple is meaningfully better than Amazon or Google on this dimension. HomePod does not retain audio recordings by default; Siri requests are processed using a random identifier rather than your Apple ID, and Apple does not use Siri audio for advertising. The privacy delta between Apple and Amazon/Google on smart speakers is significant. For households with children, HomePod’s default privacy posture is substantially more protective.

Age-Appropriate Smart Home Guidelines

The research on child development and ambient technology use points to specific age-appropriate guidelines that differ from how smart speakers are typically introduced.

For children under 6, the developmental concern is not primarily privacy but voice-assistant interaction patterns. Research by Garg et al. (2020), published in Pediatrics, found that children 3-6 who used voice assistants regularly showed reduced persistence when asking questions of human adults — they had developed an expectation of immediate, always-available, always-patient responses that did not generalize to human interaction. The privacy concern is also real at this age: children under 6 have no concept of data collection or that their voice is being recorded, and they are among the highest-frequency smart speaker users in many households.

For children ages 6-10, the appropriate approach is supervised interaction with explicit explanation of what the device records. Research supporting the broader framework of kids and online privacy shows that children who understand the basic mechanism of data collection — that the device listens, sends the audio somewhere, and stores it — make meaningfully different choices about what they say than children who believe the device is purely local. This is not a reason to frighten children; it is a reason to have a brief, accurate conversation about how the device works.

For children ages 10-14, smart speaker interaction is an entry point to conversations about the broader data ecosystem around AI tools — including what data is collected, who has access, and how it might be used. These conversations are more effective and more lasting than privacy settings alone.

What to Watch for Over the Next 3 Months

The FTC’s 2023 enforcement action against Amazon included a requirement for Amazon to submit compliance reports over a multi-year period. Watch for those compliance reports, which will be public and will indicate whether Amazon has actually implemented the deletion and retention changes it agreed to. Consumer advocacy organizations including the Electronic Privacy Information Center (EPIC) and the Center for Digital Democracy have committed to monitoring Amazon’s compliance.

Watch also for the FTC’s proposed COPPA rule update to proceed through the regulatory process. The proposed rule explicitly addresses “mixed audience” devices — general-audience devices that have child-specific features — and would potentially extend COPPA’s consent requirements to smart speakers with child modes, regardless of the general-audience classification of the device itself. If finalized, this would close the gap that allowed Amazon and Google to operate child-facing features without the full set of COPPA protections.

For parents with Amazon Echo devices specifically: Amazon’s post-settlement implementation of automatic voice history deletion was rolled out in 2023-2024 but the default is still set to “Don’t automatically delete.” Check your settings now rather than assuming the post-settlement changes apply automatically to your account — they do not in most configurations.

Frequently Asked Questions

Does Amazon’s Echo Kids change anything about child privacy?

Amazon Echo Kids is a child-specific version of the Echo Dot that is designed and marketed for children, which means it is actually covered by COPPA as a product “directed to children.” It requires parental consent, applies content filters, disables purchasing, and uses Amazon’s COPPA-compliant data handling. If you have a smart speaker in a child’s room or primarily used by a child, Echo Kids is significantly more protective than a standard Echo with Amazon Kids mode added. The key distinction is that Echo Kids is designed-for-children hardware; a standard Echo with Kids mode software is a general-audience device with an optional child profile.

Can my child’s voice be used to identify them across devices?

Voice biometric identification from smart speaker data is technically feasible. Whether Amazon, Google, or Apple currently use smart speaker voice recordings to build child voice biometric profiles is not publicly confirmed, but research in computational paralinguistics has demonstrated the capability. The FTC’s 2023 action against Amazon did not address voice biometrics specifically. This is an area of ongoing advocacy by digital rights organizations. The practical mitigation is reducing voice recording retention to the minimum available setting on your platform.

What should I tell my young child about why the speaker is listening?

Research suggests concrete, accurate explanations work better than either avoidance or alarm. “When you talk to Alexa, your voice travels to Amazon’s computers, they figure out what you asked, and they remember it for a while” is accurate and accessible to children ages 7+. You can add: “That’s why we don’t tell Alexa private things — things we wouldn’t want strangers to keep a copy of.” This is truthful, practical, and not alarmist.

Are smart TVs and gaming consoles similar data risks?

Yes, in many cases more so. Smart TVs with voice search features (available on Samsung, LG, and others) have faced FTC scrutiny for similar ambient audio collection issues. Gaming consoles with voice capabilities (Xbox Kinect, PlayStation voice features) present analogous risks. The same framework applies: these are general-audience devices that are not classified as directed to children, that children use extensively, and whose data collection is not covered by COPPA’s consent requirements. The settings audit recommended for smart speakers applies to all always-listening household devices.

What does Apple do differently with HomePod?

Apple processes Siri requests using a device-generated random identifier rather than your Apple ID, which means your voice history is not linked to your account in the same way as Amazon and Google. Apple does not sell data for advertising and has no ad-supported business model that incentivizes data retention. Apple commits to not retaining audio unless users explicitly opt in to share recordings for product improvement, and that opt-in is clearly presented. The Consumer Reports analysis of smart speaker privacy consistently ranks Apple HomePod as significantly more protective than Amazon Echo or Google Nest for all users, including children.

Should I remove smart speakers from my home entirely?

The research does not support a blanket recommendation to remove smart speakers — the question is whether you use them with appropriate settings and awareness. Families who use smart speakers for music, timers, weather, and general information with proper settings configured and children who understand the device is listening are in a meaningfully different situation than families using default configurations with no privacy settings enabled and children who interact with the device as if it were a private conversation. The privacy harm from smart speakers is real but it is substantially reducible through settings and conversation.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

Sources

1. Federal Trade Commission. (2023). FTC, DOJ charge Amazon with violating children’s privacy law by keeping kids’ Alexa voice recordings forever and undermining parents’ deletion requests. ftc.gov/news-events/news/press-releases/2023/05/ftc-doj-charge-amazon
2. Federal Trade Commission. (2023). COPPA Rule Review: Staff Report and Proposed Amendments. ftc.gov
3. Pridmore, J., & Mirmalek, Z. (2021). Smart home technologies and household data: Towards a sociology of domestic surveillance. Surveillance & Society, 19(2), 143–158.
4. Schuller, B. W., Steidl, S., Batliner, A., Hirschberg, J., Burgoon, J. K., Baird, A., Elkins, A. C., Zhang, Y., Coutinho, E., & Evanini, K. (2016). The INTERSPEECH 2016 computational paralinguistics challenge. In Proceedings of INTERSPEECH 2016. ISCA.
5. Garg, R., Cui, H., & Lindqvist, J. (2020). Alexa, are you listening to me? A longitudinal study of interactions with a voice-based AI assistant. In Proceedings of CHI 2020. ACM.
6. Edison Research & NPR. (2025). Smart Audio Report: Smart Speaker Adoption in U.S. Households. edisonresearch.com
7. Consumer Reports. (2024). Smart Speaker Privacy: How Amazon, Google, and Apple Compare. consumerreports.org
8. Electronic Privacy Information Center (EPIC). (2023). In the Matter of Amazon Alexa: Analysis of FTC Complaint and Settlement. epic.org