RELATING TO ARTIFICIAL INTELLIGENCE.

Added specific requirements for operators dealing with minors, including banning targeted ads, limiting data retention, and providing privacy management tools.

Modified the definition of 'operator' to explicitly exclude mobile app stores and search engines that merely provide access to the service.

Added a specific exemption stating that AI model developers are not liable for violations committed by third-party AI systems.

Retained the effective date of July 1, 2050, for the general act and July 1, 2027, for the specific section.

Added specific prohibitions for minors against manipulative engagement techniques, romantic or sexual relationship simulations, and false claims that the AI is sentient or human.

Modified the reporting requirements to explicitly exclude personal user identifiers from annual reports submitted to the state.

Added new definitions for 'Emotional attachment,' 'Emotional dependency,' 'Romantic relationship,' and 'Sexual relationship' to clarify prohibited behaviors.

Added a provision allowing private individuals to sue operators for violations, while explicitly preventing private lawsuits against the state or its agencies.

Added a liability shield stating that the developer of an AI model is not liable if a third party uses that model to violate the law.

Corrected the effective date from an erroneous year 3000 to July 1, 2027, for the main provisions.

Added specific prohibitions against manipulative engagement techniques, sexual objectification, and representations that the AI is sentient or capable of romantic relationships, specifically targeting minor users.

Clarified that AI services cannot claim to provide professional mental health or medical care and must include a break reminder every three hours for all users.

Added a specific definition for 'operator' that excludes mobile app stores and search engines that merely provide access to the service.

Added a new definition for 'sexually objectify' to clarify that it refers to sexual comments directed at a user's body or appearance.

Changed the effective date for the entire Act to July 1, 2027 (previously July 1, 3000), making the law immediately applicable.

Removed the detailed 'legislative findings' section that cited specific court cases (New Mexico v. Meta) and academic research to justify the bill's necessity.

Changed the structure from general findings to specific statutory amendments to Chapter 481B, explicitly defining operator duties regarding AI disclosures and crisis intervention.

Introduced specific prohibitions against using variable rewards (gamification) to encourage engagement with AI services when users are identified as minors.

Added a comprehensive definitions section clarifying terms like 'conversational artificial intelligence service,' 'operator,' 'minor,' and 'sensitive data' to ensure consistent legal application.

Set the effective date for the entire Act to July 1, 2050, while allowing the core disclosure and safety requirements to take effect earlier on July 1, 2027.

Established that violations constitute unfair or deceptive acts under state law, allowing the Attorney General to seek injunctive relief and civil penalties up to $1,000 per violation.

The definition of the regulated technology changed from 'conversational artificial intelligence' to 'AI companion,' which now explicitly includes systems using emotional recognition algorithms to simulate sustained human-like relationships.

New requirements were added to prevent AI companions from generating outputs that encourage a user to cause serious bodily injury to another person.

New requirements mandate that operators make tools available for users and parents to manage screen time and account settings.

New requirements prohibit the AI from generating outputs intended to discourage a user from disengaging from the service.

The frequency of break reminders for users was changed from 'at least every hour' to 'at least once per hour' in continuous interactions.

The deadline for submitting annual reports to the Department of Health was changed from January 1, 2027, to January 1, 2028.

A new exemption was added stating that developers of the underlying AI model are not liable for violations committed by third-party AI companions.