A coalition wrote sample AI policy so districts would not start blank

The classroom note

A principal told to 'just write an AI policy' by January often starts with a blank page and a looming deadline. TeachAI's AI Guidance for Schools Toolkit, first released in 2023 and updated in 2025, was built for exactly that moment: sample guidance language, a 'Framework for Incorporating AI,' presentation slides and communication templates, produced by a coalition the toolkit says is 'led by Code.org, ETS, the International Society for Technology in Education, Khan Academy, and the World Economic Forum' and 'advised by a diverse group of 100+ organizations, governments, and individuals.'

What the evidence says

What the toolkit itself establishes is coordination and reach, not measured effect: it is a template document assembled by a named coalition, not a study of what happens when a district adopts it. The document is explicit about its own limits, stating that 'examples and suggestions provided in this toolkit are not meant to be copied verbatim' and are meant only to 'anchor understanding in practical examples,' with each system expected to develop guidance 'in accordance with an education system's established processes.' TeachAI's homepage adds coordination detail, noting work 'in coordination with the World Economic Forum' and 'in partnership with CSTA' on computer science guidance, but neither page reports how many districts have actually adapted the sample language, or with what result.

The implementation question

The mechanism a district actually uses is adaptation labour: someone, usually an administrator with no legal or AI background, has to take the sample considerations, framework and principles and rewrite them for a specific state's data-privacy law, a specific union agreement and a specific community's concerns, none of which the toolkit can pre-solve. The toolkit is free, which lowers the barrier to starting, but the real cost is the staff time spent turning generic templates into an enforceable local policy, plus whatever legal review a district separately commissions to check the result.

What holds and what fails

What holds is that a shared starting vocabulary, produced once by a broad coalition, is more efficient than fifty states independently drafting the same categories of guidance from nothing, and the 2025 update suggests the coalition is revising the toolkit as practice matures rather than treating the 2023 version as final. What fails if a district skips the adaptation step: the toolkit's own warning against verbatim copying implies a policy pasted in unedited may not fit local law or local risk, and neither source here evaluates how many districts made that mistake.

• Which sections of our policy were adapted for local law, and which were copied from the template unchanged?
• Has legal counsel reviewed the adapted guidance against our state's specific privacy and procurement rules?
• When the toolkit is next updated, who is responsible for revising our local policy to match?

A toolkit can save a district from starting blank; it cannot substitute for the local judgement the toolkit itself says is still required.