Quick answer: The Texas Responsible AI Governance Act (HB 149, "TRAIGA") took effect January 1, 2026. It carries civil penalties up to $200,000 per violation, AG-exclusive enforcement, and a 60-day cure period. The deepfake provision targets developers of AI systems built with the sole intent of producing CSAM or unlawful deepfakes, not general-purpose generators or the platforms that distribute the output.
The Texas Responsible AI Governance Act passed the 89th Legislature in 2025, was signed by Governor Greg Abbott on June 22, 2025, and took effect on January 1, 2026. As of May 2026 it has been enforceable for about five months. The Texas Attorney General's office has been setting up the complaint intake and enforcement infrastructure, but no public enforcement action has been announced yet.
This post walks through what TRAIGA actually requires of AI developers and deployers operating in Texas, where the deepfake-specific provisions land, what the $200,000-per-violation penalty does and does not reach, and how the Texas regime interacts with the federal TAKE IT DOWN Act.
For the broader technical grounding on how AI deepfakes are produced and labeled, see the pillar guide on what a deepfake actually is.
$200,000 per violation
The civil penalty ceiling under TRAIGA for each violation by an AI developer or deployer subject to the Act. Enforcement is exclusive to the Texas Attorney General, with a mandatory 60-day cure period before an enforcement action can be brought. The penalty applies per violation, not per affected individual, which means a single AI system used in many violations can compound the exposure quickly.
Source: Latham & Watkins TRAIGA analysis; Texas Legislature HB 149 text.
What TRAIGA Actually Prohibits for Deepfakes
The law's deepfake-related provisions are narrower than the headline penalty suggests, and the framing matters. TRAIGA targets the AI system itself, not the downstream output, and not most actors who use general-purpose AI systems.
The "sole intent" provision. TRAIGA prohibits developing or distributing an AI system "with the sole intent" of producing, assisting in producing, or distributing child sexual abuse material or unlawful deepfake video or images. The phrase "sole intent" is doing significant load-bearing work. A general-purpose image or video generator that can produce unlawful deepfakes but was not built solely for that purpose falls outside this provision. The provision targets the narrow category of purpose-built abuse tooling (the "nudify" apps, the dedicated CSAM generators, the face-swap tools specifically marketed for non-consensual intimate imagery), not Stable Diffusion or Midjourney or the major foundation-model providers.
Impersonating a child in explicit text-based conversations. TRAIGA separately prohibits intentionally developing or distributing an AI system that engages in explicit text-based conversations while impersonating a child under 18. This is the conversational analog to the CSAM-generator provision and targets a separate product category (companion-AI products that allow user-defined personas).
General AI-system governance provisions. TRAIGA imposes broader obligations on AI developers and deployers around behavioral manipulation, unlawful discrimination, and constitutional-rights infringement. These provisions apply to general-purpose AI systems and create the bulk of the compliance burden the law imposes on the major AI providers operating in Texas markets.
What TRAIGA does not do. TRAIGA is not a takedown statute. It does not require social-media platforms to remove deepfakes within a specific window. It does not create a private right of action for individual victims of deepfakes (the AG is the exclusive enforcer). It does not regulate the downstream deployer who uses a general-purpose AI tool to produce an unlawful deepfake, except to the extent that deployer is itself developing or distributing AI systems. The takedown gap is filled by the federal TAKE IT DOWN Act; the deployer-side conduct is reached primarily by separate state criminal laws on harassment, CSAM, and non-consensual intimate imagery.
How TRAIGA Interacts with the Federal TAKE IT DOWN Act
The federal TAKE IT DOWN Act, signed May 2025 and enforceable May 19, 2026, creates a 48-hour platform-side removal obligation for non-consensual intimate imagery, including deepfakes. TRAIGA creates an AI-system-side civil penalty regime that took effect January 1, 2026. The two regimes layer rather than overlap.
A scenario that triggers both:
A developer in Texas builds an AI system specifically designed to produce non-consensual intimate images of identifiable women. A user employs that system to generate deepfake intimate images of a private individual. The images are posted to Instagram.
Under TRAIGA, the Texas AG can pursue the developer for distributing an AI system built with the sole intent of producing unlawful deepfakes, with civil penalties up to $200,000 per violation. Under the federal TAKE IT DOWN Act, the victim can file a takedown notice with Instagram, which must remove the content within 48 hours of receipt under the platform-side removal pathway. Federal prosecutors can separately bring criminal charges against the user for publication of a digital forgery under the criminal provisions of the same Act, on the same fact pattern that produced the first TAKE IT DOWN Act conviction in April 2026.
The Texas regime hits the supply side of the AI-system market. The federal regime hits the platform-distribution side and the individual-perpetrator side. Neither regime alone addresses the full pipeline; together they create the first end-to-end legal exposure for a deepfake operation in the US market.
What TRAIGA Means for the AI Provider Market
The major AI providers operating in Texas (the LLM vendors, the image-generation companies, the video-generation companies) face TRAIGA's broader obligations around behavioral manipulation, discrimination, and constitutional-rights infringement. The deepfake-specific "sole intent" provision is narrower and primarily affects a different market segment: the niche developers building tools specifically for unlawful synthetic content.
For comparison, the EU AI Act Article 50 transparency regime taking effect August 2, 2026 imposes machine-readable marking obligations on AI providers across all synthetic content, with penalties up to €35 million or 7% of worldwide turnover. The Texas regime is more targeted (sole-intent provision is narrow; broader provisions are focused on discrimination and manipulation) but with a $200,000-per-violation ceiling that, when compounded across many violations, can still produce meaningful aggregate exposure.
The two-track pattern is the working theory for how state and federal AI regulation will develop in parallel: the EU and forthcoming federal regimes set the broad transparency floor for AI-generated content, and state laws like TRAIGA add targeted enforcement teeth for specific categories of harm where the AG can demonstrate clear violations.
What This Means for You
The practical effects for ordinary consumers in Texas are indirect rather than direct.
Reporting still happens through the existing channels. A Texan whose likeness was used in a non-consensual deepfake reports to the platform (using the TAKE IT DOWN Act takedown notice flow), to the FBI IC3 at ic3.gov, and (where the system itself appears purpose-built for the abuse) to the Texas Attorney General's TRAIGA complaint intake. The state regime adds a third pathway; it does not replace the first two.
The detection skill is still primary. TRAIGA does not change what shows up in your feed today. The labels-vs-removal posture Meta announced for July 2026 and the Article 50 transparency obligations are the regulatory floor on the consumer-facing side. The seven Instagram detection signs, the voice-cloning listening flow, and the video-call gesture tests remain load-bearing.
The enforcement trajectory will sort over the next year. The Texas AG's office is setting up the complaint intake and staffing the enforcement operation. The first announced enforcement action under TRAIGA's deepfake provisions will establish how aggressively the office reads the "sole intent" language and which categories of AI system fall in the sights. The same pattern that played out with the first TAKE IT DOWN Act conviction is likely to repeat: the highest-visibility cases set the enforcement priorities for the longer tail.
If a deepfake video or AI-generated post caught your attention recently and you want to check whether the account behind it has been flagged by the Ledger community, paste the URL below.
Think you found an AI video?
Paste the URL and let the Ledger community verify it. Free.
The Bigger Pattern
TRAIGA is the most aggressive state-level AI governance act enacted in the US to date. The $200,000-per-violation civil penalty, the AG-exclusive enforcement, and the targeted prohibitions on purpose-built abuse tooling combine to make Texas the highest-stakes US state market for AI providers and the most targeted state for niche operators building dedicated deepfake-abuse systems. The narrow scope of the deepfake-specific provisions (sole-intent framing) is the deliberate trade-off the legislature made to avoid sweeping in general-purpose AI systems and the major providers operating those systems.
The Texas-plus-federal stack is the current US baseline: TAKE IT DOWN Act for the platform-removal and individual-perpetrator side, TRAIGA for the AI-system-developer side, the FTC and state AGs for the broader fraud surface documented in the FBI IC3 2025 report. For the EU side of the same regulatory stack, the Article 50 transparency regime takes effect August 2, 2026. Consumers operate at the intersection of all of these; none of them removes the need for the detection skill.
Related Posts
- Federal Law Now Forces Platforms to Remove Your Deepfake in 48 Hours. Here Is How to Use It.: the federal-pathway sibling that covers the platform-removal regime running parallel to the Texas AI-system regime
- The TAKE IT DOWN Act Just Got Its First Convictions. Here Is What the Enforcement Record Looks Like.: the early-enforcement-record reference for how state and federal AI-deepfake laws move from on-paper to in-court
- The EU AI Act Article 50 Deadline Hits August 2, 2026. Here Is What Changes for What You See on Platforms.: the EU-side parallel regulatory regime targeting AI provider transparency rather than the Texas system-developer enforcement model
- What Is a Deepfake? A Plain-English Guide for Social Media Users: the pillar grounding on how synthetic media is produced and why TRAIGA targets the AI-system layer rather than the output
