Texas Passes First Comprehensive AI Law in the US

The Texas legislature has passed a broad AI law that will change the face of AI regulation in the US if signed by the governor (who has until late June to do so).  If signed, the bill will become law on January 1, 2026.  Although three other states (Colorado, Utah, and California) have enacted AI laws applicable to certain kinds of AI systems or certain types of use cases, Texas' bill would regulate all manner of AI at both the development and deployment state, including use by government entities.  The bill would be enforceable by the state Attorney General, and could be used to impose fines of up to $200,000 for violations.  The law, known by the acronym “TRAIGA" (Texas Responsible Artificial Intelligence Governance Act), is drafted into and becomes part of existing laws that address data privacy and use of biometric data.  

WHY IT MATTERS

TRAIGA is important in many ways, but there are two chief aspects that merit a closer look at this stage: it has been passed at an extremely interesting time, and it is the most comprehensive effort to regulate AI in the US.  Although most small companies will not be directly affected by this law, it creates new standards that we will probably see replicated in future efforts by other lawmakers.  

As to timing: the federal government has sent mixed signals regarding AI regulation in the last year.  The changeover from the Biden to the Trump administration has been marked by several efforts to shape national regulation of cybersecurity governance and safety via executive order.  Those efforts are still taking shape, but the general approach today seems to favor less regulation and more self-governance for the AI industry as a whole.  Meanwhile, Congress has not passed any meaningful AI legislation; but the budget bill currently being fought over on Capitol Hill contains provisions that would expressly ban state-level AI laws for the next ten years.  If such language remains in the budget bill, it would override (and presumably produce a legal fight about) existing state AI laws including TRAIGA.  

As to substance: TRAIGA sets forth several express goals, including “responsible development” of AI, safety, notice, and transparency.  To address these goals, it creates a government AI Council to guide AI use and development in the state, allows developers to apply for a test permit that would protect them from liability during testing of AIs under specified conditions, empowers the state AG to investigate and punish violations of the law, and imposes limited conditions and restrictions on certain “developers” and “deployers” of AI systems.  Government entities and entities providing health care services must disclose up front that a person is interacting with an AI system.  TRAIGA bans certain use of AI by government entirely, and bans government use of biometric data without consent to uniquely identify an individual.  It also outlaws use of an AI with the intent to discriminate against a protected class, or to use an AI to manipulate humans into harming themselves or others or committing crimes.  TRAIGA also mimics a recent federal effort that outlaws creation of “deepfake” pornography.  The law applies to both generative AI (such as ChatGPT) and decision-making algorithms.  The law expressly protects consumers only in their personal or household capacity.  

Apart from the government and healthcare settings, TRAIGA does not require notice to consumers that an AI system is in use.  The prohibitions on discrimination, manipulation of human behavior, and deepfake porn, however, apply to all regulated entities.  

Violations will be reportable via an online tool to be created by the state, so that consumers can report issues.  If investigated, a regulated entity must report to the AG about the purpose, use, safety and monitoring measures employed, and other aspects of the AI that is the subject of the complaint.  

All in all, this is a broad bill that marks a giant first step for AI regulation in the US.  It is likely to face legal challenge if passed; and if Congress successfully bans state AI laws, TRAIGA may have a short lifespan.  Regardless, this is a monumental bill with the potential to set precedent and “best practices” for years to come, both in regulatory/legislative efforts and in self-governance by companies that develop or use AI.