AI in Schools
Trump softens on Anthropic, Furze pushes back on the exam panic
Digital Attitudes, 22 June 2026
The Monday read is the export control that locked Australian classrooms and lecture theatres out of Anthropic's two most capable models getting walked back at the White House inside one calendar week of the lockout, one Melbourne education writer holding the state assessment authorities to their own handbooks on what exams cannot do, and one CQUniversity adjunct pushing back on the federal primary principals' call to abolish the NAPLAN writing test for an AI enabled assessment model.
The Anthropic export freeze starts to thaw
President Trump told the Axios Show this weekend that he "might have" viewed Anthropic as a national security threat last week, but he no longer does, with the operative quote on Anthropic CEO Dario Amodei being "well, not now, but a week ago, maybe." The same interview confirms senior Anthropic staff met administration officials at the G7 tech summit in France over the previous week, with Trump telling Axios that Amodei responded to the export control directive "very quickly" and "responsibly." iTnews on the Trump Axios interview, 22 June, with Anthropic's 12 June statement on the original suspension for the full chronology.
Why this matters: For the Australian university Chief AI Officers and Chief Information Officers reading the calendar against the term 3 staffing plan that the Saturday brief priced at the Mythos 5 and Fable 5 lockout, the operational read is that the export control the IT register has been holding open as the supplier risk line for ten days is now in the active negotiation window the White House intends to close, and the question the term 3 paper answers is whether the AWS Bedrock and Microsoft Azure Anthropic procurement lines get held at the Sonnet 4.6 fallback or whether the contract template gets the Fable 5 access clause re-added in the next renewal cycle. For state Department of Education AI procurement teams running the NSWEduChat, the VicGov Microsoft tenant, and the Queensland Corella platform vendor lists against the same Anthropic dependency, the read is that the sovereign AI capability paper the December federal budget priced against a permanent foreign access risk is now sitting in front of a Trump administration that has visibly walked the threat language back, and the question the next inter-jurisdictional AI meeting answers is whether the local LLM line gets the December funding it was costed for or whether the procurement vehicle now reverts to the OpenAI and Anthropic frontier model line the universities prefer. For ed-tech vendors selling Anthropic backed AI feedback, marking, and tutoring products into Australian schools and universities (the Nurture AI feedback platform AISWA backed last week, the Cogniti platform USyd took to Microsoft Marketplace this month, the ChatMQ open source assistant at Macquarie now headed to La Trobe under Phil Laufenberg), the product roadmap read is that the model availability line in the next pitch deck has a credible path back to the most capable Anthropic tier inside the same financial year, and the vendor whose offering has already routed to Sonnet 4.6 as the production model is the one whose customers do not need a second migration in 12 months.
Furze holds the state assessment authorities to their own handbooks
Dr Leon Furze's Saturday post "Exams Are Not the Answer: Part Two" reads the AI cheating panic against the published handbooks of every Australian state and territory assessment authority, and walks the room through the QCAA Section 10 line that "external assessment is not privileged over summative internal assessment," the NESA student guidance that "school based assessment tasks can measure a wider range of skills and knowledge than an external exam can," and the syllabus level requirements in Visual Art, Drama and the practical sciences that cap exam weighting and mandate process based coursework, before naming the implementation: schools converting to wall to wall exam conditions in response to the Mazenod style AI scandals are in breach of the mandatory assessment requirements in Queensland and Western Australia. Leon Furze on Exams Are Not the Answer Part Two, 20 June.
Why this matters: For independent and Catholic school principals signing off the term 3 assessment policy under board pressure to "do something about AI," the operational read is that the QCAA, NESA, VCAA, SACE Board, SCSA, BSSS, and TASC handbook lines are now sitting in a single referenceable post the board chair will read on the same iPad as the principal's risk paper, and the question the next board meeting answers is whether the school's revised assessment policy can defend its exam weighting against the syllabus or whether the cap the regulator publishes pulls the school back to coursework moderation. For state Department of Education assessment policy teams running the post Mazenod paper for the next ministerial advisory, the read is that the Furze argument has done the cross jurisdictional comparison the policy team would normally bill out, and the question the next paper answers is whether the moderation review extends the existing safeguards or whether the department writes the new AI specific provision the Furze post argues is already covered. For ed-tech vendors selling AI proctoring and secure assessment products into Australian schools (the ProctorU Honorlock line, the Examus offering, the Inspera secure browser, the Cisco Webex room based exam product), the procurement read is that the buyer's syllabus side counsel just got a public reading list against the assumption that more exams equals more security, and the vendor whose pitch deck still anchors on "we lock the room" is the one whose next conversation has to answer the regulator's own definition of what an exam can prove.
Handwriting opinion meets the AGPPA push to scrap the NAPLAN writing test
The Educator K/12 carried Dr Ragnar Purje, Adjunct Senior Lecturer at CQUniversity, on Sunday with an opinion piece reading the late May Australian Government Primary Principals Association call to abolish the NAPLAN writing test against two decades of neuroscience on the cognitive cost of replacing handwriting with keyboards, citing the line "students who wrote their notes by hand scored significantly higher on measures of recall, knowledge, and understanding" and the working position that "the neural connectivity patterns underlying handwriting and typewriting are distinctly different," the latter the case file behind the operational claim that the AGPPA's "AI enabled" modernisation is the wrong reform to run on the writing assessment. The Educator on Purje on handwriting, 15 June, reading against The Educator on the AGPPA's call to abolish the NAPLAN writing test, 27 May.
Why this matters: For ACARA, the federal Department of Education and the Education Ministers' Meeting that owns the National Assessment Program review against the AGPPA proposal, the operational read is that the writing test debate now has a credible cognitive science side the review panel cannot answer with the "the world has moved on" line, and the question the term 3 communique answers is whether the writing assessment stays in the on screen form the Year 5 disruption flagged or whether the review reverts to the pen and paper position the Purje paper argues the data still supports. For state curriculum authorities and primary school principals reading the writing PD calendar against the existing literacy strategy, the read is that the next term 3 staff meeting now has two opposing operational frames to choose between, and the question the literacy lead's paper answers is whether the school's writing instruction gets re-anchored on the handwriting research base or whether the AI literacy line takes the curriculum time the writing instruction would otherwise carry. For ed-tech vendors selling on screen writing assessment and AI essay marking products into Australian primary schools (the WritingGrade NAPLAN tool, the Pearson Writing Coach, the Khan Academy Khanmigo writing tutor, the Eduaide Tasks rubric), the read is that the AGPPA's premise that the writing test should "modernise" against AI just got a counter that lands on the same cognitive ground the marking product is sold against, and the vendor whose next pitch still anchors on "we score the typed essay faster" is the one whose next sales call answers a literacy lead who has just read the Purje paper.
The Monday thread is that the Anthropic export control that locked Australia out of Fable and Mythos last week now sits in front of a Trump administration walking the threat language back, the Furze post that does the cross jurisdictional homework on what the state assessment handbooks already say about exam weighting, and the Purje counter to the AGPPA proposal that puts handwriting back in the writing test conversation. More on the operational read at digitalattitudes.com.au.