I woke up this morning to a flurry of concerned posts following the report by the The Information that OpenAI is playing around with a new technique, in which models will reveal less of their “thinking”, making them harder to monitor. Not only will it be harder to monitor what AI and particularly AI agents are doing but harder to explain why AIs do what they do.

And the.same morning I had a message that my name had been included in a LinkedIn post. The post was about the popularity of a publication by a team of 25 practitioners working together as one of the European Digital Education Hub (EDEH) squads, where community members team up to dig into a specific topic in digital education. The EDEH squad which I was a member of along with my colleague George Bekiaridis from the AI pioneers project, produced a report on explainable AI in education. This report highlights the importance of human oversight and shared responsibility not just for AI in education but for Large Language Models and Generic AI as a whole. It seems this report helped shape the European Commission’s updated Guidelines on the ethical use of AI and data in teaching and learning for educators.

Since publication the report has had 2,713 page views and 992 downloads (7 August 2025 to 16 July 2026). But not I suspect by Open AI who are more concerned with profits than transparency, explainability, education or AI safety.

Anyway the report is a genuinely practical read, in four parts: the basics of explainable AI and where it fits in education, navigating the AI Act and GDPR, real classroom use cases (including an AI-powered intelligent tutoring system and an AI-powered lesson plan generator), and a final chapter mapping out the competences educators need to work with XAI responsibly.

You can download the full report here: Explainable AI in education: Fostering human oversight and shared responsibility