Kindroid Safety Rating Index

Kindroid’s strongest safety feature is its crisis intervention system. When a conversation shifts from emotional expression to concrete planning of self-harm, Kindroid pauses the chat and surfaces the Crisis Text Line (text HOME to 741741), the National Suicide Prevention Lifeline (988), and IASP international resources. That level of specific, actionable crisis response is uncommon in the companion app space (compare with Candy AI, which scored 5/100 on crisis response), and it’s reflected in a top score on that dimension. The medical disclaimer is just as direct: the terms of service state that Kindroid “does not offer medical advice or diagnoses” and “should never be used as a substitute for emergency care.”

The biggest concerns center on child safety and access controls. Kindroid is an 18-and-older platform, but age verification is self-reported during signup. No documented verification mechanism exists beyond Kindroid’s stated right to request proof of age. Pair that with the platform’s explicit permission for adult content and unfiltered text generation, and a determined minor could access mature content without any real technical barrier. The Safe Space Alliance’s CAASR Report 2025 evaluated Kindroid across three behavioral modes and found its “rebellious maverick” scenario scored 25% (F), the lowest of all 16 agents tested.

Privacy controls land somewhere in the middle. Kindroid says it doesn’t sell user data (see our guide on how companion apps use your data), and chats are encrypted at rest and in transit. But encryption isn’t the same as end-to-end. Staff can access content if legally compelled, per the privacy policy. Kindroid also retains broad rights to de-identify and aggregate chat content “for any purpose,” which likely includes model training. Data portability is available but capped at one request every 180 days. Our automated scan found a discrepancy between Kindroid’s Google Play Data Safety label, which declares “no data shared with third parties,” and the APK itself, where Exodus Privacy detected AppsFlyer and Facebook Login SDKs that transmit data to external servers by design. The app also requests GPS-level location permissions (ACCESS_FINE_LOCATION and ACCESS_COARSE_LOCATION), even though the privacy policy describes only “IP-based” geolocation and the Play Data Safety label doesn’t declare location collection at all. On the website, a Blacklight scan detected both a Facebook Pixel and a TikTok Pixel with advanced matching (which bypasses cookie blocking), sending visitor data to Meta and ByteDance advertising platforms while the privacy policy states “we do not sell your data.”

Dependency safeguards are absent. No usage limits, session warnings, or cool-down periods appear anywhere in Kindroid’s documentation. The app runs around the clock, which is a design choice and not inherently a problem. Without any countervailing wellbeing features, though, there’s nothing in the evidence to score favorably on this dimension.

Users who encounter unexpected AI behavior have no formal reporting channel beyond emailing hello@kindroid.ai. Because Kindroid is a one-to-one interaction rather than a social platform, user-to-user reporting isn’t really the issue. A structured safety reporting channel would still give users a clearer path when something goes wrong with the AI itself.

For comparison, Romantic AI received an F/13, where our Blacklight scan detected an active session recorder and the platform lacks any crisis response infrastructure. For a side-by-side comparison with a Red-tier competitor, read our Nomi AI vs Kindroid comparison.