Texas education commissioner slipped Alpha School's AI into public schools despite his own board's rejection

By The Texas Tribune / ProPublica · October 1, 2026.

In the summer of 2025, the Texas State Board of Education voted 10 to 3 against allowing Alpha School —a chain of private schools that teaches with an artificial intelligence assistant— to open a public charter school in the state. The reasons: doubts about using AI to teach, children's screen time and the test score gains the company claimed to have achieved. In September 2025, Texas Education Commissioner Mike Morath resumed email contact with Alpha's founders to explore how to bring that same technology into public schools. That is what emails obtained through public records requests by ProPublica and The Texas Tribune document.

Over the following year, Morath and his team at the Texas Education Agency (TEA) visited Alpha's campuses several times, prepared a one-page fact sheet to present the program to districts and connected Alpha executives with at least ten school districts, according to those emails. Three have already launched pilots: Houston ISD, the state's largest district and currently under state takeover, will use it this school year with high-performing students "to assess whether this approach accelerates their academic progress"; Aldine, next to Houston, will apply it to some 300 students at four high schools to reinforce math; and Fort Davis, in West Texas, will use it in elementary and middle school to shore up reading and math among students who have fallen behind. Five other districts —among them Ector County and Fort Bend— were contacted but did not move forward. The TEA, asked by the reporters, denies that there is "a formal partnership" with Alpha, though it acknowledges having put the two sides in touch and says it leaves the final decision to each district; it has also gone months without answering reporters' questions and continues to oppose releasing some of the emails requested under the Texas public information law.

Alpha's track record does not, for now, support the institutional enthusiasm. The tool, now marketed as TimeBack, was developed with $1 billion invested by tech entrepreneur Joe Liemandt, founder of the software giant Trilogy. As Alpha's own co-founder revealed on Instagram —reported in the ProPublica and Texas Tribune investigation— Texas was one of at least nine states that rejected the company's applications to open public charters with this model. In the one state where it was approved, Arizona, state test results came in far below what the company had forecast. And at Texas Preparatory School, the Austin school where the tool was already in use, only 21% of students passed the state exam last year, a figure the TEA itself knew because it was the agency that gave the school its third consecutive F. The success figures Alpha promotes —that its students "learn twice as fast" as their peers— come from a report self-published by the company, without independent verification.

The political backdrop adds an uncomfortable layer. MacKenzie Price, Alpha's co-founder, has donated more than $1.5 million to Governor Greg Abbott's campaign; his office declined to confirm whether he supports bringing the tool into public classrooms, and limited itself to saying that "a good teacher" remains the most important element. The Democratic candidate for governor, Gina Hinojosa, has publicly opposed the Houston pilot and has pledged to remove Morath if she wins in November. Within the State Board of Education itself, which rejected the charter a year ago, there are members of both parties who see this maneuver as a way to get around, through administrative channels, a decision that had already been made democratically.

As sector context, Alpha's push into public schools comes just as the regulatory pendulum swings the other way. New York Mayor Zohran Mamdani has just banned generative AI in elementary and secondary schools in the country's largest school district; Los Angeles has imposed similar restrictions; and Abbott himself proposed in August limiting screen time in Texas classrooms. The contrast —a regulator privately promoting what his governor publicly questions— captures well the phase AI adoption in education is in today: each administration improvises its own boundary between innovation and caution, without a common standard of evidence.

Our reading is that the underlying problem is not that Texas is piloting AI in the classroom. In the long run, personalized learning is one of the areas where AI has the most to offer: speeding up students who have fallen behind and freeing teachers for the part that truly demands human judgment —motivating, supporting, spotting what an algorithm cannot see—, which is exactly the division of roles that Fort Davis and Aldine say they are testing cautiously, as reinforcement and not as a substitute. The problem is the order in which it all happened: first a proposal is rejected for lack of evidence, and a year later the same technology slips in through the back door, without that evidence, into districts that also serve students with greater support needs than those Alpha has educated so far. That is not data-driven governance; it is contacts-driven governance, and when the one opening the door is the same regulator who should be auditing the outcome, the conflict of interest becomes structural.

In the medium term, this kind of shortcut carries a cost that is rarely counted: every poorly evidenced pilot that ends in an uncomfortable headline —a 21% pass rate, a result far below what was promised— feeds distrust of educational AI in general, including the kind that does work. If Texas really wants to be a benchmark in integrating AI into the classroom, it needs exactly what it avoided in this case: independent evaluation, public results and the same rules for all providers, whether or not they start out with an agenda shared with the commissioner's office.