A Statement on the Responsible Use of AI in Education and Work

We build and deploy AI in classrooms and workplaces, and on the whole the people now warning against it are right to worry.

Cognitive surrender or offloading is real: when a tool can produce the answer, the temptation is strong to skip the thinking that would have produced it, and the loss does not announce itself until well after it has happened. Critical thinking, sustained attention, the patience to sit with a hard problem — these are skills and abilities built through use, not preserved by default. A student who lets AI write the essay has not learned to write, whatever the essay reads like; an employee who lets AI make the judgment call has not developed the judgment, however good the call turns out to be. We take this seriously, and it shapes how we build.

That said, the answer is NOT to withdraw from AI, or to treat every use of it as some kind of compromise. Tools have always changed what counts as skilled work and skilled learning, calculators and spellcheckers included, and the honest question was never whether to use a tool but which ones strengthen a person and which quietly hollow them out. That is the line our business tries to hold, and these are the principles behind it.

1. Augmentation: AI augments capability; it does not replace it. We design for use that extends what someone can already do, not as a substitute for learning the underlying skill in the first place. A tool that lets a competent writer draft faster is doing something quite different from a tool that writes for someone who cannot yet write, and we build for the former.

2. Productive Struggle: This is not ours to remove. Learning happens in the friction — research, retrieval, drafting badly before drafting well, getting something wrong and correcting it. Where AI would remove that friction entirely, we strive to design it out or gate it behind demonstrated competence instead, even when the smoother version would sell more easily.

3. Visible, not Disguised Use: Where AI has shaped a piece of work, that ought to be plain to whoever is assessing or relying on it. We do not build processes or features whose purpose is to make AI-assisted output indistinguishable from unassisted work, in a classroom or a workplace.

4. Judgment about People stays Human. AI can support an assessment; it should not be the assessor of record, particularly where the outcome affects someone's grade, hiring, or standing. This is a firm line for us rather than a preference, and precisely because it is a difficult one to compromise on quietly, it needs stating plainly.

5. Children and Adults are Different. A child who has not yet built foundational skills - reading fluency, arithmetic, the basic grammar of an argument - needs those built the slow way first; AI has a smaller, more supervised role here than it does for an adult working from an established base. We design age-appropriate limits into our products and services rather than leaving that judgment to the user.

6. Creativity and Originality remain the Human's to claim. AI is useful for iteration, for surfacing options, for doing the mechanical parts of a creative task - it is not the author. Where our tools touch creative or original work, they are built to support a person's voice, not to supply one in its place.

7. Adoption by Evidence, not by Fear of falling behind. We bring AI into a workflow or a curriculum because we can show it helps, not because competitors are doing it or because the technology happens to exist. That is slower, and it means turning down some features that would sell easily. It is the only honest basis to build on in this space.

None of this makes AI safe by default, in education or anywhere else. For us, it makes safety a design decision rather than an accident, one we have to keep making deliberately as the technology changes. That is the standard to which we hold ourselves, and the one we are glad to be judged against.