A maintained timeline of regulatory actions, complaints, law changes and major safety-policy shifts that affect how we score AI companion apps.
App-specific enforcement or complaints can affect an app's privacy score. Category-wide laws affect our evidence standard, but they do not punish a single app unless the source names that app.
Replika / Luka
The Italian Garante found GDPR issues around legal basis, transparency and minors. This is why Replika carries a regulatory-history penalty in our privacy scoring.
Source: European Data Protection Board summary of Italian authority actionReplika / Luka
The filing asks the FTC to investigate alleged deceptive and unfair practices. It is a complaint, not an FTC enforcement order, so we describe it carefully.
Source: Tech Justice Law Project complaint and petitionCharacter.AI
The company said under-18 users would lose open-ended character chat and move to a more restricted experience. We treat Character.AI as adults-for-chat in reviews and comparisons.
Source: Character.AI company blogCategory-wide
Covered companion chatbots face disclosure, safety-protocol and reporting duties. The law reinforces why privacy and safety are separate scoring dimensions from conversation quality.
Source: California SB 243 bill text via LegiScanCategory-wide
New York became an early US state to regulate emotionally responsive companion models. We track this as category context, not as an app-specific penalty unless an app is named in enforcement.
Source: Morrison Foerster overview of New York and California companion AI lawsWe do not treat privacy as a vibe check. We look for regulatory history, public complaints, data deletion controls, training-use disclosures, minor safety controls and whether the product encourages users to share highly sensitive personal information.
A complaint alone is not the same as a regulator finding wrongdoing. A law is not the same as an app violation. The timeline exists so the reasoning is visible and so future review updates can point to specific evidence instead of vague concern.
Related reading: our AI companion privacy checklist, data access guide, and 100-point scoring method.