AI Tools for Schools
1 storyDigital Promise opens $1 million K-12 AI Infrastructure Program for open tools that give teachers oversight of AI-enabled formative assessmentNEW
Digital Promise, working with Learning Data Insights, DrivenData, Georgetown University's Massive Data Institute and Catalyst @ Penn GSE, will fund up to two awards of as much as $600,000 each. The money is for openly licensed, shared infrastructure that helps teachers monitor AI tools, guide them and adapt instruction, and every applicant must also submit a demonstration application that engages teachers. Applications are open to US-based entities until 20 November, awards run for 18 months from 1 February 2027, and recipients are expected early 2027. This is a funding call, so there are no results yet.
Innovation lens: Open, shared infrastructure is a procurement answer as much as a technical one: it lets many schools inspect the same layer instead of each trusting a different vendor's dashboard. UNESCO's discussion paper on deliberative governance of AI in education, whose consultation closed on 15 October, treats procurement and total cost of ownership as governance levers in their own right. Leaders can ask any assessment-tool vendor what teachers can see, change and switch off, and whether another school could audit it.
Policy & Dialogue
1 storyUS coalition report: of more than 800 academic papers on AI in K-12 reviewed by Stanford's SCALE Initiative, only 20 produced strong causal evidence – and the federal research agency that would fund more has lost about 90% of its staffNEW
The Alliance for Learning Innovation, a coalition of nonprofits, researchers, practitioners, funders and private-sector leaders, has published "Investing in R&D for Safe and Effective AI in Education: Federal Policy Recommendations". It asks Congress and the administration to build a stronger research base, improve AI-enabled tools, build research infrastructure, coordinate a single federal agenda and use AI to make research funding more effective and transparent. The report sets its case against two facts: teens already use AI for schoolwork, according to nationally representative surveys from Britebound and Common Sense Media, while the evidence on what works is sparse. It notes that the Institute of Education Sciences (IES) lost about 90% of its employees in last year's staff reductions, went more than 15 months without holding research grant competitions, and has yet to see more than half of its 2026 budget released by the White House Office of Management and Budget. The report is a coalition's policy ask, and the SCALE count reflects strict criteria for causal evidence, not a verdict on every other study.
Leadership lens: Leaders are making adoption decisions on an evidence base that a strict review puts at roughly 2 to 3 studies in every 100, and the body that would add to it is shrinking. Until that changes, a school's own pilot is part of the evidence: write down the question, the comparison group and the review date before launch, so a local trial counts for more than a vendor case study. The OECD Digital Education Outlook 2026 reported a 48% task gain with AI and a 17% drop once it was removed, which is the kind of finding only well-designed comparisons can produce.
Research & Evidence
1 storyTeacher Tapp: AI use in English schools has passed 2023 predictions for lesson planning and parent communications but lags far behind for marking and dataNEW
Teacher Tapp compared what 5,254 teachers predicted in April 2023 with what 8,837 teachers reported in September 2026, both weighted to reflect teacher and school demographics. Primary senior leaders using AI for parent communications reached about 56% against 28% predicted, and primary classroom teachers using it for lesson planning reached 55% against 39%. Primary attainment-data analysis came in at about 7% against 34% predicted, and secondary marking at about 7% against 31%. Behaviour-data use sat 24 to 26 points below prediction for secondary teachers and leaders. The comparison sets forecasts from one sample against reports from another, so it shows direction rather than precise change.
Evidence lens: Teachers adopted AI where a mistake is cheap to catch and held back where a judgment about a child is involved. That is a better guide for CPD than vendor roadmaps: start training with the tasks already in use and treat marking and data as areas that need a deliberate case.
Ethics, Assessment & Academic Integrity
1 storyForbes: AI is forcing schools to explain why learning matters, and to assess the process, not just the productNEW
Columnist Ray Ravaglia argues that AI can complete assignments, so the value of an assignment lies in the reasoning and writing it develops, and take-home work cannot be reliably authenticated. He points to presentations, oral exams and discussion of revisions, and to projects with real-world stakes, such as investigating a local problem, that give students a reason to judge AI output. He cites a Duke business school professor who required recorded team meetings, analysed with transcription and an AI model, after unrestricted AI use weakened collaboration, and notes that results can be wrong and privacy protections are essential. Many of his examples come from OpenAI or its partners, and the column is opinion, not a study.
Practice lens: Recording students to prove authorship swaps one integrity problem for a surveillance problem. A short viva on a submitted piece, or a dated drafting checkpoint, gets most of the benefit without a recording. Leaders who adopt either can tell families what is captured, who sees it and when it is deleted.
Human Factors
1 storyEducation Week: how to use AI with English learners without encouraging cognitive offloadingNEW
In Larry Ferlazzo's Classroom Q&A series, teacher Kelly Cray suggests students build simple apps with AI tools toward a personal goal, make bilingual study guides and self-quiz, and use AI to adapt examples to their backgrounds while keeping the mathematics intact. Bill Tolley, at an international school in Addis Ababa, has students struggle with difficult text before turning to AI, starts each lesson with a paper journal and ends with peer discussion, uses a Socratic AI tutor, and asks students to keep an AI co-authorship journal. The post draws on research including Nataliya Kosmyna's "Your Brain on ChatGPT" preprint (reduced engagement, ownership and retention when relying only on an LLM) and Judith Kroll's work on bilingualism as a "desirable difficulty". These are teachers' practices, not tested interventions.
Language lens: For a student learning in a second language, the struggle with a text is where the language is acquired, so a tool that removes it removes the learning. The paper-journal opening and the discussion close are cheap to adopt and keep the language work with the student. Schools with many home languages can add a rule that students first try in whichever language they think in.