AI reading assistant Amira Learning has public-facing pages meant to reassure parents and teachers about data privacy, but these documents contain little more than standard cookie tracking details. This gap shows the tension as schools adopt AI voice tools while trying to comply with strict federal and state biometric privacy laws. Parents and educators want to know where children’s voice data actually goes.
What Happened
Schools across the country face pressure over AI implementation. As we previously reported, school districts in New Mexico paused the use of Amira over concerns about how student voices are processed. Amira hosts public resources to address these worries, including an official Privacy Policy, a Trust Center, and a blog post about what families should know about student data.
These pages do not explain how student speech recordings are protected or stored, or if they train external AI models. Instead, all three resources list standard website analytics and marketing trackers, like HubSpot forms and Vimeo cookies. This lack of transparency led to the pushback that occurred when districts paused the tool previously for these exact reasons.
The Bigger Picture
Voice-recognition AI tools use different methods to handle student speech. Some enterprise tools, such as Glean, route audio streams directly to AI models and do not retain raw recordings. OpenAI's voice mode discards raw audio immediately after transcription, keeping only plain-text transcripts. Other tools, like SpeakON, temporarily store voice recordings to generate transcripts, deleting them unless users consent to feedback analysis.
These differences in data storage make compliance difficult for schools. Under FERPA, schools must maintain direct control over student records shared with vendors and ensure that data is deleted when contracts expire. The Federal Trade Commission's 2025 COPPA amendments added "biometric identifiers," including voiceprints, to the definition of protected personal information for children under 13. The FTC also rejected a "school authorization exception," meaning companies cannot easily bypass parental consent requirements.
State-level biometric laws are also strict. In Illinois, the Biometric Information Privacy Act (BIPA) requires written consent before collecting voice recordings and allows private citizens to sue companies over violations. Because speech-to-text platforms process biometric data, complying with federal and state laws is difficult.
What This Means for Families
When a student reads aloud to an AI assistant, they share biometric data. School districts that fail to review software policies can violate state biometric laws and federal student privacy rights. Because many EdTech privacy policies focus on website cookies rather than voice data, parents cannot simply trust a company’s "Trust Center" badge. The design of these tools matters, as we previously reported during the rush to restrict classroom AI.
What You Can Do
- Request the specific data sharing agreement (DSA) signed between your school district and AI vendors to see how your child's voice recordings are stored and deleted.
- Ensure that any speech-based classroom app complies with local state laws like BIPA and the FTC's COPPA biometric rules.
- Ask your child’s teacher for an alternative, non-AI reading assessment if you are uncomfortable with voice-to-text software collecting biometric data.