Texas Bills Would Create Standalone State Department of AI

Texas legislators will weigh three bills expanding state control over artificial intelligence when the 90th legislative session opens January 12, 2027 — and the most ambitious one would build an entirely new state agency to police it.

Rep. Salman Bhojani, D-Euless, filed House Bill 605 to create a standalone Department of Artificial Intelligence. The department would monitor compliance with AI requirements, evaluate government AI systems, and run a public registry of covered systems used by state agencies. It would sit outside the existing Department of Information Resources and answer to a five-member board appointed by the governor with Senate consent. Board appointments must reflect expertise in AI, cybersecurity, privacy, government administration, and public procurement "to the extent practicable."

Bhojani didn't stop there. His second measure, HB 474, would require state agencies to adopt policies guaranteeing meaningful human oversight of certain AI-driven decisions. A third bill, HB 416 from Rep. Mary González, would bar state agencies from using AI systems for employment decisions — hiring, promotions, terminations and more.

All three bills were filed after bill filing opened November 9, 2026, and would take effect September 1, 2027, if enacted. More than 1,260 bills had landed by 5 p.m. that Monday.

The bottom line: Texas wants a dedicated AI cop, and it's spelling out the paperwork in advance.

What would the new department actually do?

Under HB 605, state agencies would maintain inventories of covered AI systems and submit details on their use, purpose, data categories, and safeguards. The department would run risk-based reviews, prioritizing systems subject to "heightened scrutiny" and those presenting documented risks to individual rights, government services, security, or public funds.

The agency's most public-facing product would be a searchable registry identifying covered AI systems, their governmental purposes, and whether human review is available for certain decisions. Agencies would prepare public impact summaries for systems facing heightened scrutiny. Deadlines are already written into the bill:

• Initial statewide inventory due December 1, 2028
• Public registry established by March 1, 2029

Procurement gets its own section. The department would develop a model checklist with DIR and the comptroller covering performance testing, data restrictions, vendor accountability, incident reporting, and subcontractor responsibilities. Agencies eyeing high-scrutiny AI systems would submit intended-use summaries before contracting. The department could issue written risk recommendations within 30 days — but its silence wouldn't block an otherwise lawful procurement.

The legislation preserves DIR's existing purchasing authority, AI standards, training, and implementation duties. The two departments would coordinate oversight and set up information-sharing procedures. Notably, the new department would not independently regulate private AI companies or impose duties on local governments.

In plain terms: this is oversight of government AI use, not a regime for Texas businesses.

How much human control do the other bills mandate?

HB 474 kicks in when AI systems make or serve as a controlling factor in "consequential decisions," or when they present a reasonably foreseeable risk of substantial harm to health, safety, property, critical infrastructure, or confidential information security. Oversight must come from an appropriately trained employee with authority to approve, reject, modify, override, or terminate the AI system's decision.

Texas already has governance rules on human oversight. DIR's code of ethics, effective March 2026, requires governmental entities to maintain human review capabilities and ensure AI systems can be paused, restricted, or disabled. HB 474 would layer specific statutory requirements on top, with DIR responsible for setting minimum standards including the timing and documentation of human review. DIR would have until March 1, 2028, to adopt rules, and agencies would need interim safeguards in place before then.

HB 416 takes a blunter approach. It would prohibit state agencies from using AI for employee performance reviews, hiring, promotions, demotions, terminations, and other employment decisions. It would also bar state employees from entering confidential, non-public information into AI systems. Unlike the other two measures, HB 416 bans specified uses rather than building oversight procedures.

One bill builds an agency, one mandates a human check, one draws red lines — three different regulatory philosophies in a single package.

Is Texas starting from scratch?

No. The proposals build on several laws enacted during the 89th legislative session that established the state's current AI governance framework. Senate Bill 1964 directed DIR to establish an AI code of ethics, risk management standards, and governance measures. HB 2818 created an AI division within DIR. And the Texas Responsible Artificial Intelligence Governance Act, effective January 1, 2026, set restrictions on certain AI uses, disclosure requirements, and the Texas Artificial Intelligence Council.

In an August update, DIR outlined its implementation of those requirements, including governance resources, employee training, and a public-sector program supporting evaluation and testing of AI systems. HB 605 would add a separate oversight function rather than replace DIR's existing division.

What happens next?

The bills face a divided legislature. Democrats hold the House majority after last week's elections, but Republicans retain control of the Senate. None of the three measures has been enacted, and their provisions could change during the legislative process. The regular session begins January 12, 2027, with bill filing continuing through March 12.

If HB 605 survives both chambers, Texas would move from an AI division buried in an IT department to a cabinet-level watchdog with its own board, registry, and procurement gate — a structure few states have attempted.