Free seats for teachers and scientists: where Anthropic wants Claude to be the default

Anthropic introduced Claude for Teachers on August 11 and, on August 28, made it available to US K-12 schools and districts. The same day it announced a Claude Team plan for Scientists: 10,000 scientists get access, standard seats are free, and a premium tier with 5x the limits costs $15 per month. Free is a price, and it is the most informative part of both announcements. Anthropic is not selling to teachers and scientists; it is trying to become the default tool in two professions that train the next generation of users and produce the next generation of high-value data.

What the teacher product does

According to the launch coverage, Claude for Teachers can work against the academic standards of all 50 US states, generate lesson plans, personalise materials for individual students, and work with teaching data. Those are the four jobs a teacher spends unpaid evenings on, which is why OpenAI and Google are pushing into the same classrooms. The standards alignment is the feature that matters operationally: a lesson plan that cites the wrong state standard is worthless to a teacher under evaluation, and a model that gets it right is a real hour saved per week.

What it does not change is the part of teaching that is not document production. Managing a room of thirty people, noticing the kid who stopped participating, deciding what to skip when the schedule collapses: none of that is in the feature list, and none of it should be. The honest framing is that the product moves the paperwork, not the profession.

Free seats are not generosity; they are a bet that whoever owns the lesson plan in 2026 owns the graduate in 2036.

What the science plan signals

The scientist plan is more interesting for what it sits next to. On August 27 and 28, Anthropic also put the Model Hardware Standard into research preview: a standardised driver between an operating system and physical lab equipment (robotic arms, microscopes, liquid handlers) through simple read and write commands, with the device's physical characteristics and safety limits carried in the driver rather than in a manual. It is the tool-calling pattern from MCP pointed at hardware, model-agnostic, currently limited to selected organisations, with an open-source plan.

Put the two together and the shape is obvious: free seats to get Claude onto the scientist's desk, a hardware standard to get it onto the bench. The $15 premium tier at 5x limits is where the money would eventually come from, but 10,000 seats at $15 is not a revenue line for a company at Anthropic's scale. It is a distribution line.

Pricing as an adoption signal

If you are deciding what to standardise on in a lab or a school district, read the price list this way:

• Free standard seats mean the vendor expects to make money elsewhere: on institutional contracts, on premium upgrades, or on the ecosystem. Ask which.
• A 5x premium tier at $15 tells you the standard limits are tight enough that heavy users will hit them. Find out what the base limit is before you assume free covers your workflow.
• 10,000 seats is a cap, not an adoption number. It is a pilot dressed as a programme.
• Compare to list prices. The prices page shows what the same tokens cost through the API; the gap between that and free is the size of the subsidy, and subsidies end.

What is not in the announcement

None of the coverage I have seen says how student data is handled, whether teacher prompts train future models, what "5x limits" is 5x of, or how many of the 10,000 seats have been claimed. The teacher product is US-only and aligned to US state standards, so it says nothing about European classrooms. The Model Hardware Standard is a closed preview with no published specification. All of this is adoption strategy read from a price list and a launch post, not a review of a product I have used in a classroom or a lab. The strategy is legible; the execution is not yet visible.