Claude for Teachers Is Mostly Just Claude

Anthropic announced the launch of Claude for Teachers earlier this month. The main criticism I’ve seen is that Anthropic shouldn’t be marketing directly to teachers and progress shouldn’t come at the expense of student privacy. Those concerns are legitimate, but framing the service as a choice between progress and privacy misses a more interesting set of questions: what does it actually do, how does it work, and how does it help teachers?

There’s a separate set of questions related to why hundreds of vendors market directly to teachers in the first place. The alternative is usually a tedious and slow process in which the people making the purchasing decisions are not the users of the product. District buyers also have a difficult time assessing the quality of rapidly evolving AI tools, which is why public benchmarks and evaluation frameworks will be so critical.

At the same time, data privacy concerns are real. K-12 data requires a level of care and protection that makes handing teachers unrestricted access to consumer AI tools problematic. However, this isn’t fundamentally different from any other organization that deals with sensitive data. Districts will need to figure out what their guardrails are without handcuffing teacher’s legitimate use. That is a District challenge, not a vendor one.

What’s Actually New

Claude for Teachers is designed to save teachers time in two key areas: lesson planning and differentiation. There is evidence to support the idea that AI tools can help teachers. A Stanford study cited in Anthropic’s press release concludes that “early causal studies suggest that educator-facing AI tools may reduce time spent on tasks such as lesson preparation while maintaining instructional quality.” To understand how, we need to look at what it is.

• Free access to Claude.ai, Code, and Cowork for one year with higher level usage limits for verified US K-12 teachers
• Two skills “grounded in learning science” for lesson planning and differentiation
• A connection to the Learning Commons Knowledge Graph, a preexisting tool to “integrate quality content and data directly into AI-powered tools to boost accuracy, instructional alignment, and relevance”

Additionally, the press release includes references to nine education-related platforms or services such as MagicSchool and Canva Education. Those are optional education related plugins with separate user accounts required. Those can already be used with or without Claude for Teachers. Finally, Anthropic announced new terms of service and data processing agreements specific for K-12, written with FERPA in mind.

The Skills Are the Product

Claude for Teachers is built almost entirely from existing capabilities. The skills are primarily what make Claude become Claude for Teachers. They instruct Claude how to perform lesson planning and differentiation and let it use tools such as the Knowledge Graph and code. A skill is pre-written instructions for how to perform a specific task. Claude keeps the skill name and short description in memory and loads it when the user asks. The skill itself will refer to code and tools that can handle processing inputs and outputs. The skills are open-source and include a robust evaluation framework to validate their output.

The skill format is portable. Anthropic released it as an open standard in late 2025, and OpenAI, Google, and Microsoft adopted it within months. This raises the possibility of replicating Claude for Teachers, or your own spin on it, on different platforms or using an open-weight LLM model you control. There would be differences between these platforms, but a privately hosted version presents new possibilities for addressing data privacy risks. It turns out you can have both progress and privacy.

Anthropic, through the Gates Foundation partnership, deserves credit for developing the skills and evaluation framework as open source. I’ll leave the learning science to those qualified to judge it, but the skills themselves are genuinely interesting encapsulations of teacher knowledge for planning standards-aligned lessons, differentiating across reading levels, and connecting to materials through the Learning Commons.