Weekly edition

The pitch gets bigger, the proof stays thin

13 July to 22 July 2026

The nine days to 22 July were less about one big platform launch than about the AI-in-education sales pitch meeting its own evidence problem. Anthropic joined the queue of companies offering free tools straight to teachers, but Khan Academy's own founder used the same window to admit its first AI tutor 'did not change student learning as much as many of us hoped it would.' A rural New York district's plan to put a humanoid robot in front of senior students drew as much scrutiny of its vendor as excitement about the technology. And from the Pacific, a blunt reminder that AI readiness means little without evidence and cultural grounding to match it.

Ngā tohu | Signals this week

  1. Anthropic became the latest AI company to offer free, purpose-built tools directly to teachers, joining OpenAI, Google and Microsoft in courting the classroom market.
  2. Khan Academy's own admission that its original AI tutor underperformed is a rare public acknowledgment that access to an AI tool doesn't automatically change how students learn.
  3. A rural New York school's plan to put a humanoid robot in front of students drew scrutiny over vendor ties and procurement diligence, not just novelty.
  4. Pacific regional education leaders opened a major planning meeting by naming AI, alongside climate change and labour mobility, as a force education systems must be ready for, while insisting they stay culturally grounded in doing so.

Platform moves

Anthropic launches Claude for Teachers

Anthropic introduced Claude for Teachers, giving verified US K-12 educators free access to a version of Claude wired to state academic standards and curricula such as OpenSciEd and Illustrative Mathematics. The offer includes Claude Code and Cowork for data analysis, connections to nine education platforms, and a pledge that conversations are never used to train Anthropic's models, alongside a FERPA-aligned data agreement. It follows similar free teacher offers already out from OpenAI, Google and Microsoft.

He aha te take? Why it matters for AotearoaEvery promise in this launch, FERPA compliance, US state standards, a US-only teacher verification system, is built for American schools, not ours. NZ kaiako can't sign up, and even if a similar offer arrived here, FERPA doesn't map onto the Privacy Act or the Privacy Commissioner's IPP3A guidance on indirect collection of student data. That's the actual test any AI tool needs to pass before a NZ school adopts it: not whether it's free, but whether its data terms hold up against IPP3A and whether it's wired to the New Zealand Curriculum rather than someone else's standards. Boards evaluating any classroom AI tool this term, free or paid, should be asking vendors that question directly, not assuming a well-publicised US launch means the same protections apply here.
Read the source →

Sal Khan admits Khan Academy's first AI tutor fell short, rebuilds it

Khan Academy founder Sal Khan said the original Khanmigo, launched three years ago as a standalone AI tutor, 'did not change student learning as much as many of us hoped it would,' largely because students rarely used it. The rebuilt version is now woven directly into Khan Academy's practice problems, prompting students to explain their reasoning and offering contextual feedback without giving away answers, rather than waiting for students to ask it questions first. Khan Academy says it's seeing improved performance on follow-up problems after Khanmigo interactions, though it hasn't published figures.

He aha te take? Why it matters for AotearoaThis is worth reading alongside NZCER's own finding that fewer than half of NZ primary teachers feel confident teaching responsible AI use, and 85 percent want more training. Khan's admission says the same thing from the vendor side: bolting an AI chat tool onto existing teaching doesn't automatically improve learning, and a tool has to be woven into what's already happening in the classroom to matter. For kaiako weighing up any AI tutor, including Khan Academy itself, the practical takeaway is to judge the tool by how it sits inside your existing practice and assessment, not by how capable the chatbot sounds in a demo.
Read the source →

Classroom practice

Rural New York school partners with robot maker for a humanoid classroom assistant

Salamanca City Central School District is deploying a Realbotix M-Series humanoid robot alongside an AI tutoring assistant, Optio, for senior STEM classes and take-home tutoring this school year. Superintendent Mark Beehler said the district wants to give students 'a safe, Salamanca-specific AI tutor' and staff 'customised AI tools,' with district oversight and safeguards against inappropriate responses built in. The announcement has drawn scrutiny in US media over Realbotix's parent company also making adult products, and over whether a US$57,590 contract undercuts the case that the robot isn't there to replace teachers.

He aha te take? Why it matters for AotearoaNo NZ school is close to a humanoid robot in the classroom, but the governance question underneath this story is exactly the one boards here already face with any AI vendor: who is actually behind the product, what happens to student data, and does 'it won't replace teachers' hold up once the contract is signed. Under IPP3A, a board evaluating any AI tool, robot or otherwise, needs to know who owns the company, where student interactions are processed, and what recourse exists if something goes wrong, before procurement, not after a controversy breaks. RTLB and learning support staff should also note the caution here about embodied AI and vulnerable learners: a friendly robot interface changes the relationship in ways a chat window doesn't, and that deserves its own risk conversation.
Read the source →

Sector news

Pacific education leaders told to build AI-ready, culturally grounded systems

Opening the 11th Pacific Board for Educational Quality meeting, the Pacific Community's Dr Andrew Jones told regional education leaders that AI and digital transformation are 'changing how knowledge is created, accessed, and applied' across the Pacific, alongside climate change and labour mobility. He called for education systems that are resilient, evidence-informed, culturally grounded and future-ready, with values kept central even as new technology is adopted. The Board's meeting also reviewed the Pacific Regional Education Framework, data systems and qualifications frameworks supporting labour mobility across the region.

He aha te take? Why it matters for AotearoaThis is squarely Aotearoa's neighbourhood, and the framing Dr Jones used, resilient, evidence-informed, culturally grounded, future-ready, is close kin to what Te Tiriti-based practice already asks of AI adoption here: that new technology serve existing values rather than override them. NZ's own AI-in-education work, from Ministry guidance to individual school policies, rarely gets put in the same frame as Pacific nations we share a region with and often share learners with, through RTLB caseloads, Pacific bilingual units and cross-Pacific staffing. Schools with significant Pacific learner populations have a genuine opening here: connecting AI policy kōrero to this regional framework, rather than only to US or European guidance, grounds it in something closer to home.
Read the source →

What to do with this

School leadersBefore adopting any AI tool, free or paid, ask the vendor two direct questions: who owns the company, and do the data terms hold up against the Privacy Commissioner's IPP3A guidance. A slick US launch or a headline-grabbing robot answers neither question for you.
KaiakoIf you're trialling an AI tutor with your class, judge it by how well it sits inside your existing teaching, not by how impressive it looks in a demo. Khan Academy's own founder just admitted a standalone AI tool changed less than expected because students didn't reach for it unprompted.
RTLB and learning supportEmbodied AI, a robot or an avatar with a face, changes the relationship with a learner in ways a chat window doesn't. If any AI tool with a persona or physical presence is proposed for use with vulnerable ākonga, treat it as its own risk conversation, not an extension of existing chatbot policy.
Boards and whānauAsk your school whether its AI approach is grounded in your own community's values, not just borrowed from overseas guidance. Pacific education leaders this fortnight put culturally grounded practice on equal footing with AI readiness, which is a fair bar for any NZ board to measure its own approach against.

Browse past editions →

On the radar at home

ERO report on AI in schools Still not published as at 22 July. Sector feedback was delivered in early July; publication remains imminent.
Senior secondary AI curriculum Year 11 to 13 AI content is scheduled to become mandatory between 2028 and 2030. Consultation timeline to watch.
Ministry guidance on generative AI The operative guidance on assessment authenticity: design not detection, and generative AI remains banned in external NCEA assessment.
Privacy Commissioner IPP3A guidance Final guidance (December 2025) is now the operative document for any tool collecting student information indirectly.
Deepfake Digital Harm and Exploitation Bill Before Parliament; would expand the definition of 'intimate visual recording' to cover AI-generated and altered images.