Our February collection of articles and announcements to share this month.
Education
As we embrace ‘efficient’ AI we risk adding more work on top
As colleges race to embed AI and meet new inspection expectations, the sector may be underestimating the implementation burden. This piece argues that unless leaders actively remove other pressures, AI will add complexity instead of improving teaching quality.
Business schools search for clear AI guidelines
Business schools are racing to embed AI into teaching, assessment and operations — but shared standards are lagging behind. While innovation is widespread, leaders are grappling with how to benchmark progress, manage ethical risks and avoid superficial adoption.
Some Australian schools are embedding AI directly into classroom practice, with chatbots questioning students about their work to test understanding and deter plagiarism. But a new paper warns that uneven rollout risks creating a “two-speed” education system, with better-resourced schools pulling ahead.
Security
Anthropic says it ‘cannot in good conscience’ allow Pentagon to remove AI checks
Anthropic says it won’t “in good conscience” strip safety protections from its AI model to satisfy the Pentagon, risking a lucrative contract in the process. The row highlights growing tensions between AI companies positioning themselves as safety-first and governments seeking more expansive military applications.
US AI giant accuses Chinese rivals of mass data theft
Anthropic has accused three Chinese AI firms – DeepSeek, Moonshot AI and MiniMax – of using “distillation” techniques to extract capabilities from its Claude chatbot at scale. The company claims millions of interactions and thousands of fake accounts were used to replicate its model’s strengths, raising concerns about intellectual property theft and national security.
Assessment
What generative AI reveals about assessment reform in higher education
This HEPI blog argues that generative AI hasn’t broken assessment in higher education – it has exposed long-standing weaknesses. Rather than doubling down on detection and compliance, Dr Emma Ransome calls for a deeper rethink of how and why we assess, with trust, clarity and meaningful learning at the center.
Perceptions of AI
AI is developing so fast it is becoming hard to measure, experts say
A new chart from research group METR suggests AI systems are doubling in capability roughly every seven months – and the pace may be accelerating. But while the trend looks dramatic, experts caution that measuring AI progress is becoming harder, and the real-world impact on jobs and automation remains uncertain.
Now the struggle is no longer real, are students becoming stupid?
Are students becoming less capable, or are universities failing to adapt? Using experiments, brain studies, and policy debates, Dickinson argues that AI doesn’t inevitably make students “stupid” – but poorly designed teaching might. The article challenges higher education to rethink its purpose in an age where knowledge is always on tap.
AI ‘slop’ is transforming social media – and a backlash is brewing
AI-generated “slop” – low-quality, often bizarre images and videos designed for quick engagement – is flooding social media feeds, racking up millions of views despite obvious flaws. While platforms like Meta and YouTube are leaning into AI creation tools, a growing backlash from users suggests frustration is mounting over what many see as digital pollution.
AI shatters the pretence that academic polish was ever anything but gatekeeping
This post suggests that higher education has long conflated intellectual quality with mastery of academic register, disadvantaging working-class, first-generation and non-native English-speaking students. With journals now accepting AI for language and drafting support, McKenzie argues universities should rethink assessments and focus on the substance of ideas rather than linguistic polish. It’s a challenge to traditional notions of academic integrity in the AI age.
Vendor news
Introducing Perplexity Computer
Perplexity has launched Perplexity Computer, a general-purpose AI system designed to run complete workflows rather than just answer questions. It breaks down goals into tasks, assigns sub-agents, and orchestrates multiple leading AI models to deliver work asynchronously. The announcement signals a shift from chat-based AI tools to autonomous digital workers.
Anthropic Drops Flagship Safety Pledge
Anthropic has scrapped a key promise in its Responsible Scaling Policy that previously barred it from training powerful AI models unless safety measures were guaranteed in advance. The company says the fast-moving AI race and lack of global regulation make unilateral pauses impractical, opting instead for more transparency and regular risk reporting.
Build with Nano Banana 2, our best image generation and editing model
Nano Banana 2 is Google’s latest image-generation model, promising richer detail, more accurate text in images and faster results. It can adapt visuals for different languages and create more context-aware scenes using real-world references.
Anthropic’s launch of AI legal tool hits shares in European data companies
Shares in several major UK and European data and publishing companies fell steeply after Anthropic unveiled an AI tool designed to automate legal workflows. While the company stresses it doesn’t replace lawyers, markets reacted to the possibility that AI could erode demand for traditional legal information and analytics services. It’s another sign that AI competition is intensifying across professional services.
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