Maryland's New Law Makes Districts Prove an AI Tool Works Before Buying It
Maryland's AI Ready Schools Act forces districts to vet AI tools against a state rubric before purchase, a procurement-first model UK academy trusts don't yet have.
A different kind of AI law
Most state and national responses to classroom AI have followed the same script this year: mandate a policy document, set a deadline, hope districts comply. Ohio made policies compulsory and then had to chase stragglers past the deadline. Four more US states have told districts to write a policy without saying what a good one looks like.
Maryland has tried something narrower and, on paper, more useful. Governor Wes Moore signed Senate Bill 720, the Artificial Intelligence Ready Schools Act, on 26 May 2026, and it took effect on 1 June. Rather than asking schools to write a policy from scratch, it puts the state education department in the business of vetting the tools before they reach a classroom at all.
What the law actually requires
The Maryland State Department of Education (MSDE) must publish a rubric and evaluative tools that local school systems use to assess any AI product before it is piloted or purchased. According to legislative documents and MSDE's own guidance materials, that rubric now sits alongside procurement rules issued by the state's Department of Information Technology, so a school cannot simply sign a vendor contract because a sales rep demoed it well.
The act also creates the Maryland AI Education Collaborative, a standing body meant to keep the guidance current as tools change, and it requires AI literacy to be folded into the state's workforce-readiness and computer science standards by 1 June 2027. Some drafts of the legislation went further, proposing university-backed certification for vendors whose tools meet the state's bar, effectively an external stamp of approval before a product can be marketed into Maryland classrooms.
The mechanism worth noticing isn't the AI literacy requirement. It's that Maryland has put the burden of proof on the vendor and the procurement office, not the classroom teacher.
Why this matters more than another policy mandate
England has been here before, just from a different angle. The Department for Education's £23m EdTech Testbeds programme is trying to build an evidence base for what works, but it operates alongside, not ahead of, the market: schools can and do buy tools with no testbed data behind them. Ofsted and the DfE have published guidance on AI use, and KCSIE 2026 adds safeguarding rules from September, but nothing in England's framework stops a multi-academy trust from adopting a tool with zero independent evaluation.
Maryland's model inverts that order: evaluation happens at the procurement gate, using a single state-issued rubric, before a tool reaches a pilot. That is closer to how medical devices or textbooks get approved than how most edtech is bought today, where a free tier and a persuasive pitch are often enough. It will not catch everything, a rubric is only as good as the criteria it tests for, and MSDE has not published data yet on how many products have failed the assessment or been withdrawn. Vendors could also treat the rubric as a compliance checkbox rather than a genuine bar, the way GDPR cookie banners became a formality rather than a protection.
What UK academy trusts and MATs can take from this
The lesson for English trusts is not "copy Maryland's rubric", legal frameworks differ, and MSDE's structure relies on state procurement law that has no UK equivalent. The lesson is procedural: a rubric applied consistently before purchase is a cheaper and faster safeguard than writing a use policy after a tool is already embedded in lessons. Most English schools currently do the opposite, a teacher or head of department adopts a tool informally, and only later does the trust ask what data it collects or what evidence supports its claims.
Trusts that want to borrow the idea without waiting for national guidance could build a one-page internal rubric covering data handling, evidence of impact, accessibility, and exit terms, and require any AI tool to clear it before procurement sign-off, not after a term of use.
What to do
- Ask your MAT or governing board whether AI tools are vetted before purchase or only after complaints arise.
- Build a short internal evaluation checklist covering data protection, evidence of efficacy, and cost of exit, modelled loosely on Maryland's rubric approach.
- Watch for MSDE publishing rubric outcomes data later in 2026, if it shows vendors failing or withdrawing products, that is evidence procurement-first regulation can work, not just look good on paper.
What to watch
- Whether Maryland publishes results from its rubric, showing which AI tools passed, failed, or were withdrawn.
- Whether other US states copy the procurement-first model rather than the policy-mandate model that has dominated 2026 so far.
- Whether England's DfE or Ofsted moves any evaluation requirement earlier in the procurement chain, rather than leaving it to the Testbeds programme alone.