A standard that reads a class roster is not a standard for AI safety

The classroom note

When a school buys an AI tutoring or writing tool and asks it to know which students are in which class, or to post a grade back to the gradebook automatically, the answer depends on a set of technical rules a teacher never sees: 1EdTech's interoperability standards. Two of them do the practical work. LTI lets a learning management system launch an outside tool inside a course and, in the consortium's own words, reduces teacher workload by providing a way for tools to pass assignments and grades back to the teacher's centralized gradebook. OneRoster separately manages the roster itself, class rosters including student, teacher, organization, course, and related enrollment information, alongside gradebook results.

What the evidence says

1EdTech's own standards catalogue lists more than a dozen specifications beyond these two, covering assessment data exchange (QTI), learning analytics (Caliper), competency frameworks (CASE), and portable achievement records (CLR), each governing a distinct kind of data rather than a single unified AI safety check. The same page describes a separate Data Privacy standard used for vetting educational applications to ensure that a minimum standard of privacy and security is met, and names this as part of the organisation's TrustEd Apps certification track, distinct from LTI or OneRoster interoperability certification. That structure means a product can hold interoperability certification, proving it correctly reads rosters and writes grades, without having gone through the separate privacy, security, accessibility, or generative-AI-specific rubrics at all, because these are administered as separate reviews.

The implementation question

For a district evaluating an AI product, ‘1EdTech certified’ is not a single claim; it is shorthand that could mean the product passed a technical conformance test for one specification, several, or an entirely separate trust rubric. A vendor's marketing may not distinguish which certification it holds. The practical mechanism worth naming plainly: LTI and OneRoster certification tests whether software correctly implements a data-exchange protocol, a question about format and transport, not about whether the underlying AI model handles student data responsibly, moderates content safely, or produces accurate output. Those questions sit under the separate rubrics, including one 1EdTech now runs specifically for generative AI data handling.

What holds and what fails

Interoperability certification holds reliably for its narrow purpose: a certified LTI or OneRoster integration should exchange roster and grade data in a predictable, tested format, reducing the custom integration work a district's IT staff would otherwise face. It fails as a proxy for safety, privacy, or pedagogical quality, because those are governed by separate standards that a product may or may not also hold. Treating 1EdTech certified as a blanket safety assurance would be an editorial overreach the standards body's own documents do not support; the certificate answers a plumbing question, not a trust question.

• Which specific 1EdTech certification does this product hold: LTI, OneRoster, a TrustEd Apps privacy rubric, or several?
• Does passing interoperability certification tell a district anything about how the product's AI component uses student data?
• Is the certification current, and does it cover the version of the product actually being purchased?

A standards body can guarantee that data moves correctly between systems; it cannot, through that same certificate, guarantee what a vendor does with the data once it arrives.