Once viewed as a speculative threat to learning integrity, AI now reshaping how lessons are planned, research is conducted, and administrative tasks are handled. Two recent developments — a Welsh government-commissioned report on AI in schools and the University of Oxford’s rollout of ChatGPT Edu — illustrate the education sector’s evolving relationship with this technology. Together, they reveal a sector eager to harness AI’s benefits, but equally aware of the risks it introduces to pedagogy, privacy, and trust.
In Wales, education inspectorate Estyn found that schools are experimenting with AI to “substantially cut teacher workloads” and support pupil learning, but that adoption remains “ad hoc” and uneven. Teachers are using AI for lesson planning, drafting parent letters, writing reports, and generating tailored learning materials. Estyn reported that this has reduced administrative burdens, freeing staff to spend more time on direct teaching. At the same time, educators expressed concerns about plagiarism, bias, and overreliance on AI tools.
Birchgrove Comprehensive School in Swansea offers a case study in cautious optimism. Digital and innovation lead Ryan Cresswell described the school’s approach as “very positive” — teaching pupils to use AI responsibly rather than prohibiting it outright. Students use generative tools to summarise revision notes, create quizzes, and clarify complex topics. “If the pupils are going to be using it, we’d rather teach them how to use it responsibly than just ignore it,” Cresswell told the BBC.
Headteacher Andrew Owen echoed that sentiment, noting that AI helps some students “sift the information that’s really important,” particularly those who struggle with cognitive load. Yet he also acknowledged that “the use of AI has moved on very quickly” and that teachers are “struggling to catch up” — a concern mirrored in Estyn’s call for national guidance and staff training. The Welsh government accepted those recommendations, emphasising the need to “balance effective use with the safety and wellbeing of learners and staff.”
In higher education, the University of Oxford has taken a more structured, institutional approach. Following a year-long pilot involving 750 students and staff, Oxford has become the first UK university to offer ChatGPT Edu — a privacy-focused version of OpenAI’s platform — to all students and employees. The initiative forms part of a five-year partnership with OpenAI and is framed as part of the university’s “digital transformation.”
Professor Anne Trefethen, Oxford’s Pro-Vice-Chancellor for Digital, called the rollout “an exciting step” that could “accelerate high-impact, curiosity-led research and innovation.” OpenAI’s international education lead, Jayna Devani, said Oxford was “setting a new standard” for responsible AI in academia. The platform provides enhanced privacy controls and institution-level data security, addressing one of the key barriers that has slowed adoption elsewhere in the public sector.
Taken together, these two cases highlight a growing divide in institutional readiness. Schools are improvising — guided by digitally confident teachers rather than national policy — while universities like Oxford are formalising partnerships and developing frameworks for responsible use. The difference reflects the uneven diffusion of AI literacy and infrastructure across the education system.
Estyn’s findings show that most Welsh schools are still in the exploratory phase, where enthusiasm coexists with anxiety. Teachers’ apprehension about accuracy, bias, and safeguarding echoes wider public debates about generative AI’s reliability. Meanwhile, Oxford’s move represents the next phase: integrating AI into institutional workflows under controlled conditions, with ethical usage and critical thinking embedded into training.
A clear trend emerges: AI is being adopted not to replace educators but to redefine their focus. In schools, automation of administrative tasks allows teachers to concentrate on creative pedagogy and pastoral care. In universities, tools like ChatGPT Edu are framed as cognitive amplifiers — supporting personalised learning and research synthesis. Both contexts, however, face a shared tension between empowerment and dependency. When students or staff rely too heavily on generative tools, the risk shifts from inefficiency to intellectual complacency.
Another underlying dynamic is trust — both in the technology and in its governance. The Welsh government’s acceptance of Estyn’s call for national guidance acknowledges that individual schools cannot regulate AI use alone. Oxford’s deployment of a secure, education-specific model reflects a similar logic: confidence in AI requires public institutions to retain control over data and application contexts. These developments suggest that the public sector’s role is evolving from passive user to active curator — setting standards for safe, equitable AI integration rather than leaving it to market forces.
The UK’s education system is becoming a microcosm of the broader societal challenge of AI governance. The Welsh example shows the potential and pitfalls of bottom-up experimentation — innovation driven by motivated teachers but constrained by inconsistent policy. The Oxford case illustrates the benefits of top-down coordination, where formal partnerships and data safeguards enable broader adoption without compromising ethical standards.
If these approaches converge, the public education sector could define a uniquely human-centred model of AI adoption — one that prioritises literacy, responsibility, and wellbeing alongside efficiency and innovation. That will depend on three factors:
- Clear national guidance and training so educators can use AI confidently and critically.
- Robust privacy and safety frameworks that ensure student data remains under institutional control.
- Curricular integration that treats AI as both a learning tool and a subject of ethical inquiry.
The next phase will test whether public institutions can move beyond isolated pilots to system-wide strategies that scale responsibly. The direction is promising: schools like Birchgrove are showing how AI can augment human teaching, and universities like Oxford are demonstrating how structured governance can make that sustainable. But as both cases reveal, technology alone is not the transformation — it is how the public sector chooses to guide, constrain, and cultivate its use that will determine AI’s lasting impact on education.