Talkie AI Safety Rating Index

Talkie AI earned a D/30/Red rating across our 23-dimension safety analysis, completed in March 2026. Three sub-dimensions scored 5/100: crisis response, age verification, and safety reporting. Those failures are not theoretical. TorHoerman Law is actively investigating Talkie AI for suicide and self-harm risks (March 2026), and the original Talkie app was removed from the US Apple App Store in December 2024. Neither Apple nor developer SUBSUP PTE. LTD. (a Singapore subsidiary of Chinese AI company MiniMax) has disclosed the specific violations.

Child safety is the most urgent concern. Talkie sets a 16+ age requirement in its Terms of Service, but verification relies entirely on self-reporting. Reddit threads and app store reviews document children as young as 9 using the platform. One Reddit post describes an 11-year-old accessing NSFW content after downloading Talkie from an in-game ad. A password-protected Teenager Mode exists (it disables search and enforces nighttime downtime), but multiple user reports indicate mature content remains accessible despite it. The iOS replacement app (Talkie Lab) now carries an 18+ rating with “Frequent/Intense” mature content warnings.

Privacy practices show a troubling gap between disclosures. Talkie’s privacy policy explicitly lists collection of messages, voice content, location data, and device identifiers. The Google Play Data Safety section, however, declares “No data collected.” That direct contradiction means users who check the Play Store before downloading get a fundamentally inaccurate picture of what Talkie actually collects. The privacy policy also covers CCPA and Singapore’s PDPA but omits GDPR entirely, despite Talkie operating globally with 50M+ Google Play downloads.

Automated technical analysis revealed some of the starkest findings. Exodus Privacy detected 32 tracker SDKs in the Android build, including 21 advertising networks. That count is the second-highest in our registry (behind Chai AI’s 34) and explains the relentless ad load users complain about. One of those SDKs is SuperAwesome, a kid-targeted advertising network. Embedding a child-focused ad SDK in an app with documented child safety failures raises serious questions about audience targeting. The app also requests 52 total permissions, including calendar read/write access, approximate location, and the ability to monitor other app usage on your device (PACKAGE_USAGE_STATS), none of which have a documented purpose in an AI chat app.

Monetization practices amplify the risk. Free users face ads between nearly every interaction, chat limits, and a gem-based premium currency system with gacha mechanics. The iOS App Store listing confirms loot boxes. In an emotional companionship product with a documented young user base, that combination of scarcity mechanics and interruptive ads creates monetization pressure on vulnerable users during emotionally invested moments. The combination of character roleplay and emotional investment also raises emotional dependency concerns, particularly for younger users still developing social skills.

Talkie does get a few things right. AI nature transparency and therapeutic claims avoidance both scored 53/100: the app markets itself clearly as AI, and it does not make wellness claims. User control dimensions (data portability, conversation management, privacy settings) each scored 53/100, reflecting basic CCPA opt-outs and account deletion capabilities. For comparison, Romantic AI received an F/13, where we found a FullStory session recorder capturing every tap and keystroke. Talkie’s issues are different in kind: the problem isn’t covert surveillance, but a massive gap between what the platform claims and what it actually does to protect its youngest users.