Our June collection of articles and announcements to share this month.
Education
In the AI age, oracy can be the most valuable graduate skill – if we instil it
With AI reshaping traditional assessment, this article argues that graduates will need to prove not just what they know, but how well they can communicate it. It’s a timely discussion of how universities might rethink teaching and assessment to better prepare students for the workplace.
AI can help providers read and act on the student feedback they never usually get to
Rather than replacing human judgement, AI can help universities make better use of qualitative student feedback that would otherwise be overlooked. The authors suggest that combining AI analysis with human oversight allows institutions to identify recurring issues sooner and act before the next survey cycle.
Defining and developing students’ critical AI literacy
As AI becomes a routine part of student life, higher education needs to shift from policing AI use to developing informed judgement. This article explores practical ways to build critical AI literacy through transparency, reflection and discussions about the ethical, environmental and social impacts of AI.
Security
Anthropic’s Claude Fable 5 and Mythos 5 AI suspended over security fears
Anthropic’s latest AI model was taken offline after US officials expressed concerns over its security risks, despite the company arguing the identified vulnerabilities were limited. The decision raises wider questions about AI governance, global collaboration, and access to increasingly capable systems.
Research
Humans outperform AI at this highly rigorous mathematics test
Researchers have introduced one of the toughest AI maths evaluations yet, using previously unseen research problems and formal expert grading. The results show that AI is making impressive progress but is not yet capable of consistently matching professional mathematicians on novel research tasks.
Major UK research funders green-light AI for processing grant bids
UKRI, Wellcome and other major UK funders are softening their stance on AI by allowing its use in parts of the grants process, while ruling out fully automated funding decisions. The article also highlights early data showing that around one in four applicants to some funding schemes are already disclosing AI use.
New AI-powered train monitoring system could help reduce overcrowding on UK railways
Researchers at Loughborough University and TrainFX have developed an AI-powered system that estimates train carriage occupancy in real time using privacy-friendly depth-sensing cameras. The technology could help rail operators manage overcrowding more effectively and, in future, allow passengers to see which carriages have space before boarding.
Government
Introducing the AI Economics Institute
A new government-backed AI Economics Institute will examine how AI is affecting the UK economy, from labour markets to business productivity. Rather than setting policy, it will generate research and data to help government make better-informed decisions as AI adoption accelerates.
Academic Integrity
Universities must help shut down the illicit AI detection economy
As AI detection tools become more common, students are increasingly paying third-party services to check and lower their AI-likelihood scores before submission. The authors warn that this “AI detection economy” is shifting students’ focus away from improving their writing and towards gaming detection systems.
Environment
AI and environmental sustainability in post-16 education
Jisc’s new report explores the often-overlooked environmental impact of AI in further and higher education. It offers practical guidance to help institutions balance AI adoption with sustainability goals, even as the evidence continues to evolve.
Experts: You have some power to reduce your AI environmental footprint
As AI becomes embedded in more everyday tools, experts are raising concerns about its growing environmental impact. The article explores how data centres consume increasing amounts of electricity and water, while arguing that users can reduce their footprint by being more selective about when they use generative AI.
Environmental researchers know the impact of AI – so why do they still use it?
Why do environmentally minded researchers continue to use resource-intensive AI? Drawing on research from the University of Exeter, this piece explores how responsibility for AI’s environmental footprint is often shifted to institutions, technology providers or policymakers, and makes the case for universities to lead on sustainable AI governance.
Perceptions of AI
Stanford was their golden ticket – could AI help or hinder that?
Stanford graduates are entering one of the most AI-driven job markets in history, and many are feeling conflicted about what comes next. While some see AI as an exciting tool for innovation, others worry about its impact on jobs, education, ethics and the environment.
Authors, reviewers and editors should not be left to endure AI anxiety alone
As generative AI becomes part of the research process, existing publishing policies are struggling to keep pace. The article calls for publishers to provide practical support, consistent guidance and better resources instead of leaving reviewers and editors to manage the challenges alone.
Vendor news
Claude Fable 5 and Claude Mythos 5
Anthropic has introduced Claude Fable 5, its most capable generally available AI model to date, alongside the more restricted Claude Mythos 5 for trusted cybersecurity and life sciences users. The announcement focuses not only on improved performance but also on new safeguards designed to limit misuse while making frontier AI more widely accessible.
After a temporary pause caused by US export restrictions, Claude Fable 5 is returning worldwide with updated safety measures designed to reduce cybersecurity risks. The announcement also highlights Anthropic’s push for shared standards on AI jailbreaks and stronger partnerships with government and industry.
Previewing GPT-5.6 Sol: a next-generation model
OpenAI has announced a limited preview of its new GPT-5.6 model family: Sol, Terra, and Luna. The release focuses on stronger reasoning, coding and cybersecurity capabilities, alongside a more robust safety framework and a phased rollout before wider availability.
Microsoft explores DeepSeek for Copilot Cowork
As AI coding and workplace agents become more compute-intensive, Microsoft is moving away from unlimited pricing for Copilot Cowork. The company is exploring alternative models, including DeepSeek, to reduce costs while maintaining enterprise security and giving customers more model choice.
Dreaming: Better memory for a more helpful ChatGPT
ChatGPT’s memory is getting smarter, with a new system that automatically synthesises information from previous conversations to provide more relevant and current responses. The article explains how this should improve long-term projects, personalised recommendations and continuity across chats.
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