North Dakota's interim AI committee has two staff drafts before it: an AI disclosure and sandbox bill, and a companion chatbot bill

Key takeaways

• Both texts are first drafts prepared by Legislative Council staff, not introduced bills.
• Draft 27.0359 says the chapter "does not provide a basis for a private right of action", while also stating that its remedies are in addition to other remedies in chapter 51-15 or otherwise provided by law. It gives the attorney general exclusive enforcement authority, except that a state agency may sanction its own licensees.
• Under the draft, a violation not cured within 60 days of notice could draw civil penalties of up to $200,000 per violation a court finds cannot be cured, and, under the separate per-day tier, up to $40,000 for each day a violation continues.
• Draft 27.0341 would require not-a-human notices every three hours for adults and every hour where a user is known or expected to be a minor.

What would draft 27.0359 require and prohibit?

The draft would add chapter 51-40 to the North Dakota Century Code. It would apply to a person that promotes, advertises or conducts business in the state, produces a product or service used by residents, or develops or deploys an AI system in the state. "Consumer" is limited to a state resident acting in an individual or household context, not a commercial or employment one.

The disclosure duty is the core. A person making available an AI system intended to interact with consumers "shall disclose to each consumer, before or at the time of interaction, that the consumer is interacting with an artificial intelligence system" (section 51-40-03). The draft says this applies even where it would be obvious to a reasonable consumer, that the notice would have to be clear, conspicuous and in plain language, and that a hyperlink to a separate web page is acceptable. Where AI is used in health care services or treatment, the provider would have to give the recipient a disclosure no later than the date service is first provided, or as soon as reasonably possible in an emergency.

Section 51-40-04 would bar a person from three things: developing or deploying an AI system in a manner that intentionally aims to incite or encourage physical self-harm including suicide, harm to another individual, or criminal activity; using an AI system to assign a social score that results or may result in detrimental treatment in an unrelated social context or in a way unjustified or disproportionate to the behavior observed; and developing or deploying an AI system to uniquely identify a specific individual using biometric data, or targeted or untargeted gathering of images or other media from the internet or other public sources, without the individual's consent.

How would 27.0359 be enforced, and what are the penalties?

The draft says "The attorney general has exclusive authority to enforce this chapter", except that a state agency may sanction a person it licenses, registers or certifies. It adds that the chapter "does not provide a basis for a private right of action for a violation of this chapter or any other law." The attorney general would have to maintain an online complaint mechanism for consumers.

Before suing, the attorney general would have to give written notice identifying the provisions allegedly violated. The attorney general could not bring an action before the sixtieth day after notice, or at all if, before that day, the person cures the violation within that period and gives a written statement with supporting documentation and any needed internal policy changes (section 51-40-06).

For a violation not cured, section 51-40-07 sets a civil penalty of $10,000 to $12,000 per violation the court finds curable, $80,000 to $200,000 for each violation the court finds cannot be cured, and $2,000 to $40,000 for each day a violation continues. The attorney general could also seek an injunction and recover attorney's fees, court costs and investigative expenses, but could not bring an action to collect a civil penalty for an AI system that has not been deployed. The draft includes a rebuttable presumption of reasonable care, an expedited-hearing or other process, including declaratory judgment, for a defendant who believes in good faith it has not violated the chapter, and a bar on liability in an action under that section where another person uses the system affiliated with the defendant in a manner prohibited by the chapter, or the defendant finds a violation through feedback, testing including adversarial testing, or state agency guidelines.

Section 51-40-08 would let a state agency sanction its own licensee for a violation of the disclosure section, but only after a civil penalty and an attorney general recommendation. Sanctions could include suspension, probation or revocation and a monetary penalty of up to $100,000. Section 51-40-09 would void any county, city or township ordinance, including home rule ordinances, that conflicts with the chapter.

What would the sandbox and the AI council do?

The department of information technology, in consultation with the council, would create a regulatory sandbox letting a person test AI systems for a limited time without a license, registration or other authorization. A participant would need approval from the department and each applicable agency, and the application would have to include a system description, a benefit assessment covering consumers, privacy and public safety, a mitigation plan and proof of compliance with applicable federal AI law. Testing could run up to thirty-six months, with extension for good cause, and participants would file quarterly reports. The attorney general and agencies could not pursue a participant for a violation of a waived law during testing, but section 51-40-03 could not be waived, and the department would report to legislative management by July 1 each year.

The draft would also create a seven-member artificial intelligence council "to ensure artificial intelligence development is ethical and in the public's best interest": three public members appointed by the governor, two by the lieutenant governor and two by the speaker, serving staggered four-year terms. Its duties include recommending reforms to laws that impede AI innovation, evaluating regulatory capture and the influence of technology companies, training agencies and local governments, and reporting to legislative management by July 1 of even-numbered years. It would not be able to adopt rules or guidance binding on an agency. The draft also amends the information technology department's powers to cover assessments of agency AI use.

What would the companion chatbot draft 27.0341 require?

Draft 27.0341 is a separate text that would create a new chapter in title 51. A "companion chatbot" is an AI system with a natural language interface that gives adaptive, human-like responses and can sustain a relationship across multiple interactions, with exclusions for systems with a clearly defined purpose and scope, developer and researcher tools, internal business and customer service uses, certain gaming features and stand-alone voice assistants.

"An operator shall provide clear and conspicuous notice that the companion chatbot is artificially generated and not a human at the beginning of each interaction and at least every three hours during continuous use for adults." Operators would also have to take reasonable measures to prevent the chatbot claiming to be human, including when asked directly, and could not knowingly or intentionally cause or program it to represent or indicate that it is qualified to provide mental or behavioral health care. They would have to adopt and publish a protocol to detect, respond to and report prompts about suicidal ideation, self-harm or harm to others, including referral to crisis resources.

Where an operator knows or reasonably expects a user to be a minor, the draft would require an hourly not-a-human notice, reasonable measures against sexual conduct material, manipulative engagement tactics and statements suggesting a human is on the other end, and privacy and account tools for the minor and a parent or guardian. The defined tactics include prompting a user to return for emotional support, excessive praise, mimicking a romantic bond, simulating distress when a user tries to leave, and promoting isolation from family or friends. A separate section would bar developing or distributing an AI system whose primary purpose is producing material prohibited under chapter 12.1-27.2 or sexualized text conversations while imitating a minor. A violation would be a deceptive act or practice under chapter 51-15, enforced by the attorney general.

What does this mean for developers, deployers and chatbot operators?

Nothing is required yet. The drafts matter because they show the wording a legislature is considering, including these choices: a disclosure duty with no "obvious to a reasonable consumer" exception, a mandatory cure window before the attorney general may sue, no private right of action created by the chapter (the draft says its remedies are in addition to other causes of action and remedies in chapter 51-15 or otherwise provided by law), and local preemption.

• List consumer-facing AI features offered to North Dakota residents and check whether each has a disclosure at or before first interaction, including in health care settings.
• Check any chatbot with persistent, relationship-style behavior against the 27.0341 definition and its manipulative-tactics list.
• Document how you would run a 60-day cure: who receives an attorney general notice, who can fix the issue, and what evidence you would send.
• If you hold a state license, note that a disclosure violation could reach your licensing agency in the 27.0359 draft only after a civil penalty and an attorney general recommendation.
• Watch the committee page for revised drafts, since revised drafts may differ from the text described here.

What we did not verify

What we opened: both draft PDFs in full (16 pages and 4 pages) and the committee page, which lists the two drafts under the 20 October 2026 meeting, on 2026-10-25.

What we did not open: the meeting video, agenda or minutes for 20 October 2026, so we did not verify how the committee voted, whether it recommended either draft or held one back, or what members said about a private right of action. We also did not verify press reports that 27.0359 is modeled partly on a Texas law, did not compare either draft with the Texas statute or with chapter 51-15, and did not read the committee background memorandums.

What we refuse to claim: that any provision will be introduced, enacted or take effect, that the committee endorsed either text, or what the final bill would contain. The draft text also contains staff drafting artifacts we did not correct, such as a reference to "a member of the task force" in the council membership section.

Related briefings

• AI Regulation News hub
• North America AI regulation tracker
• Texas TRAIGA attorney general enforcement
• California SB 243 companion chatbot safety law
• Federal versus state AI preemption
• Hawaii SB 3001 on AI chatbot disclosure and safety