Week of August 3, 2026
🔴 Must Know
The EU AI Act’s high-risk deadline arrived — and then moved. 2 August 2026 was the original date on which high-risk obligations under the EU AI Act were to bite, and education sits squarely in Annex III: admissions, assessment, proctoring, AI-detection used in assessment, and systems that steer learning pathways. Under the Digital Omnibus political agreement, stand-alone Annex III obligations are now deferred to 2 December 2027 (2 August 2028 for AI embedded in regulated products). What has not moved: the AI literacy duty on every organisation deploying AI, in force since February 2025, and the outright ban on systems inferring student emotion from facial, vocal or biometric cues. For schools in the EU — including international schools operating there — the practical read is that the compliance clock has been extended, not stopped, and the banned-practice line is already live.
The AI vendors are competing for classrooms, and the entry price is zero. Anthropic launched Claude for Teachers on 14 July, free for verified US educators and mapped to state academic standards, with a Detroit Public Schools pilot studying educator wellbeing. Google and OpenAI are running parallel education plays. Free tiers arrive in schools through individual teachers rather than through procurement, which means most Heads will find these tools already in use before any DPIA, contract or data-residency question has been asked. That gap is where the governance risk lives, in every jurisdiction.
Subscribe to continue reading
Subscribe to get access to the rest of this post and other subscriber-only content.