Children’s companion chatbot accounts would need consent from their parents

Companion chatbot platforms would require parental consent for covered children’s accounts and provide parental controls. These rules would cover account holders age 17 or younger whom the platform knows or reasonably believes are Florida residents. Residents must have lived in Florida for more than six of the past 12 months. Companion chatbots sustain personalized social conversations; educational tools and certain limited-purpose chatbots are excluded. Consenting parents could obtain all conversations, limit use, block contact with other account holders, and receive alerts when a child expresses a desire or intent to harm themselves or others. Platforms would disclose that the chatbot is AI, give break reminders, and take reasonable steps to prevent material harmful to minors. Accounts categorized as belonging to minors for content or advertising targeting would face closure without parental consent, with 90 days to dispute closure. Platforms would delete closed accounts’ personal information unless another law requires keeping it. Knowing or reckless violations could bring state penalties up to $50,000 per violation and damages up to $10,000 in lawsuits on a minor’s behalf, plus legal costs. A consistent pattern of knowing or reckless failures could bring additional punitive damages. The failed proposal specified July 1, 2026. Reminders would appear at the start and at least hourly. Closure requests would take effect within five business days for minors or 10 for consenting parents. Lawsuits would have a one-year deadline from discovering, or reasonably being expected to discover, a violation.

Schools would notify parents about AI learning tools, offer an opt-out, and arrange access to account information. The rules would cover public schools, private schools, school districts, and voluntary prekindergarten providers. Parent notice, opt-out, and account access would apply to minor students. Covered learning tools serve educational purposes and are not designed to provide companionship or encourage continued use through human-like relationships. Before grade six, use would be allowed when school staff direct and supervise it, for needed support for an identified English language learner, or for needed support for a student with a documented disability. Public schools would provide an alternative activity meeting a comparable learning requirement without penalty when a parent opts out. Operators could offer read-only account access or provide information already kept by the tool after a written request. Schools and operators would not have to create or retain additional conversation records. Existing rights to education records and state and federal student privacy duties would remain in place. The failed proposal specified July 1, 2026. Notice would precede student login access. Read-only parental access would accompany student access; requested information would be available within 30 days of a written request.

AI companies would restrict personal-data sharing, bot operators would disclose automated conversations, and Florida’s Department of Legal Affairs could investigate violations. Data rules would cover companies that develop or use AI or collect data for AI, and users who live or have their permanent home in Florida. Bot rules would cover automated online software offered to Florida users. The proposal’s separate list of AI rights expressly would not create independent rights. Federal law could specifically authorize sharing identifiable information. Otherwise, companies would remove reasonable links to people and their devices, protect that separation, and require recipients to follow the same rules. Attempts to identify people would be allowed only to test the removal process. Bot notices would not apply to bots used solely by employees for a business’s internal operations. The Department of Legal Affairs alone could enforce the bot and data rules, seeking up to $50,000 per violation plus legal costs. Across these rules and the minor-chatbot rules, it could allow 45 calendar days to fix violations. A satisfactory fix with proof would block its action for that violation. The minor-chatbot fix period would not apply when a platform willfully or knowingly disregarded an account holder’s age. The department could deny future fix periods through a guidance letter. The failed proposal specified July 1, 2026. Bot notices would come at the start and at least hourly during an interaction.

Government agencies would restrict AI contracts with vendors tied to foreign countries that Florida law labels countries of concern. The restrictions would cover government purchases of AI technology, software, or products, including AI offered as one part or an option within a larger contract. Agencies could not knowingly enter AI contracts with vendors owned or controlled by a government of a country of concern, or organized under that country’s laws or mainly based there. Before accepting a bid, proposal, or reply, or entering a covered contract, an agency would need a vendor officer’s or representative’s sworn statement that the vendor does not meet those criteria. The failed proposal specified July 1, 2026, for the sworn-statement requirement and the ban on extending or renewing contracts with vendors meeting the listed foreign-country criteria.

People publicly using someone’s AI-generated name or likeness for business or advertising would need consent, with listed exceptions. The consent requirement would cover public commercial or advertising use of AI-generated names, images, and other likenesses, including those of deceased people. Express oral or written consent could come from the person or a representative authorized in writing. For a deceased person, an authorized representative could consent; without one, a surviving spouse or child could consent. A parent or guardian could act for a minor surviving child. Exceptions would cover certain genuine news uses, resale following consent to the original use, and photos of unidentified members of the public. The AI-specific news exception would require a genuine report that clearly acknowledges uncertainty about the material’s authenticity. People entitled to consent could seek a court order stopping unauthorized use and damages. Unauthorized use of a servicemember’s likeness could also bring a penalty up to $1,000 per commercial transaction, unless an exception applies. The failed proposal specified July 1, 2026. Claims under this section would not cover uses occurring more than 40 years after the person’s death.