In my last post, I proposed five principles for a new era of openness. That piece made an argument about the industry. This one is about us.
If we're asking the education community to raise its collective bar, we should say where we stand against our own principles. Some of this work is well along. Some of it isn't finished, and we’ll be transparent about where that's true. Internally we organize it under three pillars, which are interoperability, transparency, and enabling innovation. The five principles below cut across all three.
Principle #1: Expansive data interoperability
Institutions don't have one integration need. They have several, with different volumes, different sensitivities, and different costs. Treating them as a single problem is how integration strategies get expensive and brittle. So we've built four distinct levels of access, and we're pushing each one toward tighter scoping.
REST APIs. Canvas has a large, documented API surface, and it stays central to how partners and institutions build. What's changing is our attention to how those APIs actually get used. We're working toward better default access controls for the specific task at hand rather than granted broadly by default. Narrow scope helps on three fronts at once. It costs less to run, it reduces risk, and it gives an institution a clear answer when someone asks what a given tool can reach.
Data Access Platform (DAP). Some questions shouldn't be answered by hammering an API. Pulling data for analysis, reporting, or modeling is a different job that deserves a lower-load path. Through DAP–our centralized data repository and analytics backend–we support bulk access to Canvas, Catalog, Commons, and Studio data today, and we're extending that access across our platform. The same scoping principle applies, for the same reasons.
Live Events. When a system needs to react the moment something happens, listening beats pulling. Live Events lets institutions and partners subscribe to activity as it occurs rather than continuously querying to ask if activity has happened, and we offer it at no cost. A district shouldn't have to choose between an efficient architecture and an affordable one. The current event catalog covers the actions our customers use most, and we keep adding to it.
Standards-based integration. LTI, OneRoster, and other edtech standards are the connective tissue of this industry. Our support for them is free, and it will stay that way. Standards are what let a small district with a two-person IT team run a real ecosystem, and they carry security and privacy protections with them by design.
The set of companies integrating with our platform has grown too, and we're building for that. It changes the identity question in our own systems. An AI agent acting for a learner needs a defined identity and a defined scope, not an unlabeled request that looks like a person. We will be thinking through this use case as part of our continued centralized identity work.
Principle #2: Security and privacy by default
Scoping isn't only an efficiency story. It's also the security story.
Every broad grant of access is a surface, and the speed at which weaknesses are found has changed. So we treat right-sized access as a control rather than a nice to have, and we test what we publish against tools that move faster than any human reviewer.
Governance is the other half. Data governance isn't compliance paperwork that happens after a contract closes. It's a design input. Institutions that can see and control what data leaves their systems, and under what consent, are institutions that can say yes to more. We'd rather build those frameworks with our customers than hand them a policy after the fact. This has become a key discussion area for our Security Advisory Board, launched in July of this year.
Principle #3: Rules for systems moving through systems
This is the newest work and the least finished, but we are thinking about this meaningfully.
Agents are becoming users of educational systems. New protocols don't replace what this industry already built. LTI, OneRoster, and Caliper are an "and," not an "or." Work that's already open and well described is work an agent can reason about. Work that isn't open stays invisible to agents, which is the strongest practical case for open standards I've heard in years.
Some of what comes next is technical. Agents need semantic understanding of a course, not just access to its records, which is part of why we built the Ignite Search API. Some of it is governance: delegated authorization, consent, and audit trails that capture what acted and on whose behalf. And some of it can't be solved by any one company, which is why we're active in the 1edtech community as the consortium works through how MCP should behave in education. If every vendor ships its own approach without talking to anyone, we'll end up with a market full of integrations that don't do the job of integration.
One thing we're not doing is racing toward maximum autonomy. There's a real difference between what these systems can do and what we want them to do in a classroom. Keeping a person in genuine authority sometimes costs a little speed. In education, that's a trade worth making on purpose, especially with the continued resources and budget challenges facing leaders across the industry.
Principle #4: Transparency about AI use and outcomes
We continue to stand by two commitments.
The first is transparency about how AI works inside our products. Where IgniteAI touches a learner's work, educators and learners should be able to tell. Where an educator makes the final call, that should be visible in the workflow rather than buried in documentation. If AI shapes a grade, the learner should know and should be able to question it. We’re expecting our integrated Partners to adopt similar philosophies and are providing transparency about the AI incorporated in Partner tools with our AI Nutrition Facts documentation.
The second is transparency about outcomes. Our industry has a habit of describing what a product does and calling that proof. It isn't. Openness about evidence means publishing what changed, for whom, and under what conditions, including the cases where the answer came in weaker than we'd hoped. We've asked the field to do this for years and we are deeply committed to leading this work via our expert research team.
Principle #5: Translation as the new domain
Translation is where the data captured in our systems earns its keep.
We maintain human-validated crosswalks across state, national, and international standards frameworks. It's unglamorous work that makes everything downstream possible. The harder and more interesting problem is connecting those crosswalks to what learners actually do, and then to the credentials they earn. That connection is what turns a standards map into something education-aware reasoning models can work with.
Translation only holds up if the underlying records stay authoritative. Mastery holds standards-aligned assessment evidence. Parchment holds the verified record of learning. Canvas holds the coursework. Mapping across them is useful precisely because each one stays the source of truth for its own piece.
That's the discipline we're holding ourselves to. Think deeply about translation, so we can deliver against the important principle, and honor which system owns which truth.
What we're asking of you
Openness has never been a feature we finished. It's a set of commitments we renew as the ground shifts, and it's shifting quickly. We invite institutions, edtech partners, and standards organizations across the global learning community to join us in shaping this next era openness.
