Anthropic's Claude for Teachers: Free, US-Only, and Agentic by Default

Anthropic launched Claude for Teachers on 14 July, offering verified US K-12 educators free access to its premium Claude Pro tier, standards-aligned lesson tools and a set of built-in 'teaching skills'. It joins ChatGPT for Teachers, Google's classroom tools and Khan Academy's Khanmigo in what Chalkbeat has called an AI 'arms race' for America's classrooms.

Most coverage has focused on the price tag: nothing. But the detail that matters for practice is buried further down the announcement. Claude for Teachers is agentic by default, and that is a genuinely new kind of tool entering schools, not just a rebadged chatbot.

What's actually new

According to Anthropic, a teacher can hand Claude a folder of exit tickets, attendance records and class notes at the end of the school day, set a recurring task to run automatically at 4pm each afternoon, and receive a synthesised summary of what each pupil mastered, plus a proposed adaptation for the next lesson — without opening the tool again or writing a fresh prompt.

That is a meaningful shift. Every AI tool reviewed by NeuralClass so far, from MagicSchool to Khanmigo to the various grading assistants, has required a teacher to actively prompt it, review the output, and decide what to keep. Claude for Teachers is designed to work in the background, on a schedule, on live pupil data, and hand back conclusions the teacher didn't ask for at that moment.

The platform also connects to nine third-party classroom tools at launch, including ASSISTments, Brisk Teaching and Canva Education, so the same agent can potentially pull data from multiple systems rather than sitting inside one app.

The privacy trade-off

Anthropic says conversations and files submitted through Claude for Teachers are not used to train its models, and that student information is covered by a K-12 Data Processing Addendum built around US FERPA rules. Access is free for any teacher who signs up before 30 June 2027, matching the free window OpenAI already offers for ChatGPT for Teachers.

The moment an AI tool is scheduled to run unattended on real pupil records — rather than waiting for a teacher to paste something in — the question stops being whether the output is accurate and becomes who is watching what it does with the data.

That question is easier to answer in the US, where FERPA gives a defined legal frame for pupil records and a DPA that vendors can write to. It is murkier here. UK schools operate under the Data Protection Act 2018 and UK GDPR, which require a lawful basis, a data protection impact assessment, and clear limits on automated decision-making that affects individuals. An autonomous daily agent producing lesson adaptations for a child, generated without a human in the loop at the point of generation, sits close to the kind of automated processing that UK GDPR treats with particular caution — even if a teacher reviews the output before acting on it.

No UK version yet

Claude for Teachers is explicitly US-only at launch, tied to state academic standards and piloted with Detroit Public Schools for a study on educator workload and wellbeing. There is no announced UK schools product, and Anthropic's general-purpose Claude still carries an under-18 age restriction that keeps most pupils off the platform directly.

That gives UK schools a rare position: watching a live agentic rollout in another jurisdiction before it reaches them. The pattern with Khanmigo, MagicSchool and ChatGPT for Teachers has been the same — US launch first, UK adoption via individual teachers and multi-academy trusts months later, usually before formal DfE or school-level policy catches up. Ofsted and the DfE have said little specifically about agentic, scheduled AI processing of pupil data, as distinct from teacher-prompted use, which is the gap this launch exposes.

What to do

• If staff start using Claude, ChatGPT or similar tools with scheduled or recurring tasks on pupil data, treat it as a new category in your AI policy, not a variant of chatbot use — it needs its own data protection impact assessment.
• Ask any vendor pitching agentic features directly: what runs without a human triggering it, on what data, and who can see the output.
• Keep pupil-identifiable data out of any tool that isn't covered by a data processing agreement your school or trust has actually reviewed, whatever the platform's general privacy claims.

What to watch

• Whether Anthropic, OpenAI or Google announce a UK or EU version of their teacher agent products, and what data terms come with it.
• How the Detroit Public Schools pilot reports on educator wellbeing — early evidence on whether agentic tools reduce or simply relocate teacher workload.
• Whether the ICO or DfE issues guidance on automated, scheduled AI processing of pupil records, given UK GDPR's stricter stance on automated decision-making than FERPA's.