Our May collection of articles and announcements to share this month.
Education
What UK university AI policies actually do: A study of 96 institutions
A new HEPI study of 96 UK university AI policies finds that 41% of institutions have no publicly accessible AI policy, while many existing policies prioritise compliance and misconduct processes over learning. The study finds that policies often present themselves as educational while operating primarily as mechanisms for monitoring and regulating student behaviour.
90pc of students using AI for revision – but they’re ‘doing it wrong’
New research suggests that 90% of GCSE and A-level students are using AI to support their revision, with many reporting reduced stress and improved understanding of difficult topics. Experts argue that AI is most effective when used for self-testing, planning and reinforcing learning rather than simply generating answers.
Student AI use is being driven by deteriorating student-staff relationships
Drawing on conversations with undergraduate and postgraduate students, this article found that AI is frequently seen as more accessible, responsive and non-judgemental than academic support. The piece highlights how growing pressures on staff and weakening student-staff connections may be encouraging greater reliance on AI tools.
Controlled experimentation with AI – the power of a pilot
Rather than treating AI adoption as a race to implement the latest tools, this article advocates for a more deliberate, evidence-based approach. Through case studies from Exeter and Oxford, it demonstrates how pilots can help universities assess impact, address concerns around governance and data, and prepare for wider rollout where appropriate.
Academic Integrity
I knew my writing students were using AI. Their confessions led to a powerful teaching moment
MIT lecturer Micah Nathan reflects on discovering that some fiction students were using AI to write workshop submissions, leading to an unexpectedly candid classroom discussion. He suggests that overreliance on AI risks bypassing the difficult but essential work through which people develop ideas, judgement and their own voice.
Accessibility
The Web Is Being Made Accessible for AI, Not People
The article explores the overlap and important differences between AI-friendly design and accessibility. While some AI-driven infrastructure changes may bring benefits, the authors caution against “accessibility-washing,” where companies claim accessibility gains while overlooking the rights and needs of disabled people.
Research
An AI solution to an 80-year-old problem has shocked mathematicians
OpenAI has reportedly found a counterexample to a mathematical conjecture proposed by Paul Erdos in 1946, solving a longstanding problem about arranging points on a plane. The result has impressed leading mathematicians and is being described as one of the first genuinely significant mathematical discoveries produced largely autonomously by an AI system.
Security
Project Glasswing: An initial update
Anthropic shares early results from Project Glasswing, a collaboration focused on securing critical software using advanced AI systems. The findings point to a future where AI can rapidly strengthen cyber defences, but also underline the need for faster patching processes, better support for open-source maintainers, and stronger industry-wide security practices.
Perceptions of AI
A new King’s College London survey suggests many university students are worried about AI’s impact on jobs, with one in three believing rapid job losses could lead to social unrest. Despite these concerns, students remain heavy users of AI and most would still choose to attend university.
AI isn’t taking graduates’ jobs, but it will reshape entry-level roles
New data from the Institute of Student Employers suggests AI is more likely to reshape graduate jobs than replace them. Employers expect routine tasks to be automated, increasing the importance of skills such as critical thinking, judgement, adaptability and communication.
Pope Leo denounces ‘culture of power’ driving rise of AI
In a significant statement on technology, Pope Leo has called for the “disarming” of AI, urging greater oversight of its development and use. The encyclical highlights concerns about inequality, labour displacement and the growing influence of large technology firms over digital infrastructure and data.
Vendor news
Google has unveiled Gemini Omni, a new multimodal AI model that combines Gemini’s reasoning capabilities with the ability to generate and edit content across different media types, starting with video. The launch highlights a move towards more conversational video creation, where users can refine and transform clips through natural language prompts.
Making it easier to understand how content was created and edited
Google has announced a major push for content provenance, expanding AI watermarking and verification tools across its products. The move includes new detection capabilities for organisations and broader industry partnerships designed to improve trust in digital media.
Verify OpenAI-generated images
Following their collaboration with Google, OpenAI has released a research-preview tool that can check whether an image contains provenance signals linked to its image generation systems. It looks for C2PA Content Credentials and SynthID watermarks, helping users identify images that may have originated from ChatGPT, the OpenAI API, or Codex.
Find out more by visiting our Artificial Intelligence page to view publications and resources, join us for events and discover what AI has to offer through our range of interactive online demos.
For regular updates from the team sign up to our mailing list.
Get in touch with the team directly at AI@jisc.ac.uk