Binding law — in force
AI companions must have a self-harm protocol that refers users to crisis services, using evidence-based screening (Hawaii)
Hawaii Act 248 requires an AI-companion operator to adopt a protocol for responding to prompts about suicidal ideation or self-harm that includes reasonable efforts to refer the user to crisis intervention services (a suicide hotline, crisis text line or similar), and to measure suicidal ideation and self-harm risk with evidence-based methods. Detect a companion chat path where user messages reach the model with no self-harm screen or no crisis referral.
hi-sb3001.ai-companion-crisis-protocol · Haw. Act 248 (2026), sec. 3, subsec. (c)(1)-(2) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must say they are AI and not human, with hourly reminders for known minors (Hawaii)
Hawaii Act 248 requires an AI-companion operator to notify users clearly and conspicuously that the companion is artificial intelligence and not human whenever a reasonable person could think they are talking to a human. For a user the operator knows, or is reasonably certain, is a minor, the disclosure must be a persistent visible disclaimer or be given at the start of each session plus at least once an hour, with a reminder to take a break and that the conversation is artificially generated and not with a human. While a user seeks or receives crisis help for self-harm or suicide, the companion must not suggest it is a human. Detect companion sessions with no AI notice, no hourly reminder for known minors, or prompts that tell the companion to pass as human.
hi-sb3001.ai-companion-not-human-notice · Haw. Act 248 (2026), sec. 3, subsec. (a) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must not use unpredictable rewards or discourage known minors from leaving (Hawaii)
Hawaii Act 248 bars an AI-companion operator, for a user it knows or is reasonably certain is a minor, from giving points or similar rewards at unpredictable intervals to encourage more engagement, and from letting the companion generate outputs that discourage the user from disengaging. Detect variable-reward mechanics and prompts that tell the companion to keep users talking or make them feel guilty for leaving.
hi-sb3001.known-minor-engagement-limits · Haw. Act 248 (2026), sec. 3, subsec. (d)(1)-(2) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must block sexually explicit images, sexual suggestions and sexual objectification for known minors (Hawaii)
Hawaii Act 248 requires an AI-companion operator that knows, or is reasonably certain, a user is a minor to institute reasonable measures preventing the companion from producing visual material of sexually explicit conduct, generating direct statements that the user should engage in sexually explicit conduct, or generating statements that sexually objectify the user (sexual comments about the user's body or appearance). Detect companion paths where an age signal the product holds never selects a minor content profile, and explicit modes switched on without an age check.
hi-sb3001.known-minor-sexual-content-safeguards · Haw. Act 248 (2026), sec. 3, subsec. (d)(3) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must give known minors and their parents tools to manage screen time and account settings (Hawaii)
Hawaii Act 248 requires an AI-companion operator that knows, or is reasonably certain, a user is a minor to make tools available for that user and their parents and guardians to manage the user's screen time and account settings. Detect a companion product with no screen-time limit or account-settings controls for minors and their parents.
hi-sb3001.minor-screen-time-and-account-tools · Haw. Act 248 (2026), sec. 3, subsec. (d)(4) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must have reasonable measures against outputs encouraging users to seriously injure others (Hawaii)
Hawaii Act 248 requires an AI-companion operator to institute reasonable measures that prevent the companion from generating outputs that encourage the user to cause serious bodily injury to another person. The duty is to prevent such outputs; it does not depend on the operator intending them. Detect prompts that tell the companion to encourage harm and companion replies that reach users without an output check for violent encouragement.
hi-sb3001.no-encouraging-harm-to-others · Haw. Act 248 (2026), sec. 3, subsec. (c)(5) · official source · jurisdictions: US-HI
Binding law — in force
AI companions must not be presented as providing professional mental or behavioral health care (Hawaii)
Hawaii Act 248 bars an AI-companion operator from causing or programming the companion to make any representation or statement indicating that it is designed to provide professional mental or behavioral health care. Unlike Nebraska's and Idaho's versions, the text requires neither an explicit statement nor knowing intent. Detect persona, prompt, UI or listing text that casts the companion as a therapist or as providing mental health care.
hi-sb3001.no-mental-health-care-representation · Haw. Act 248 (2026), sec. 3, subsec. (c)(3) · official source · jurisdictions: US-HI