Universities in England asked to test more work without AI
On 11 November 2026, the Office for Students, which regulates universities in England, published a report on AI with Advance HE. The regulator warns that AI threatens how students learn, and how universities test them. It wants far more work assessed without AI.
What happened
The report is called Making AI work for students. Advance HE, a body that works with universities on teaching, wrote it with the regulator. Around 300 students and staff, from more than 120 universities and colleges, took part between June and September 2026.
Students described classmates who finished marked work with AI writing tools, sometimes in just a few minutes. Those classmates got similar grades to students who did the work themselves.
The report says a significant share of university assessments are still unsupervised, so students can use AI while they work. Some can be finished without showing the skills they are meant to test.
The regulator now expects every assessment to measure what a student knows and can do, reliably. In many cases, it says, that will mean secure assessment without AI. That could be an exam in a room, a presentation, a spoken exam or a practical task.
It stops short of making that compulsory. Josh Fleming, the regulator's Director of Strategy and Delivery, called AI “a generational risk to the system as it stands”. He asked for far more assessment without AI.
The report also wants students to build AI skills. It calls for greater AI literacy: knowing how to use the technology well and ethically.
Why it matters for your child
A child who is 11 today could start university in about seven years. If universities follow the regulator, more of their marks will come from work done in a room, in front of someone.
So your child will need two things at once. They will need to use AI well, and to show what they know without it.
The report warns that some uses of AI replace the effort that deep learning needs. When a tool does the thinking, less of the learning sticks. How a child uses AI matters as much as whether they use it.
A child who understands how AI works is in a stronger place. They can use it on purpose, check what comes back, and still know the subject themselves.
What Robocoders build
In AI Story Studio, Robocoders aged 11 to 17 write their own AI app in Python. It turns a topic into a narrated video. The course starts from a blank file, so no coding is needed first.
They connect the app to a real AI model, keep its key safe and write the templates that shape the story. The app makes a picture for every scene and records the narration with text to speech. Then they join pictures and voice into a video, in code.
The course ends with an assessment of the app they built. A child who can only type a prompt is using someone else's judgement. A child who can build the tool, test it and correct it is directing it.
The first taste of AI can come before any code. From age 10, on Shadow Runner, Robocoders build a robot that reacts to light on its own. The course calls it the first example of AI: a machine that senses and decides, with no code at all.
Some Robocoders already sit real exams in a room. Our Solihull centre is a registered exam centre for GCSE Computer Science. GCSE Computer Science sat by 30+ Robocoders, the youngest aged 10.
Every Robocode class is in person, in a Studio Lab, with a Mentor in the room.
Where to start
Start with a Free Tech Assessment at your nearest centre. Your child builds and codes a first project with a Mentor. It is free, and there is no obligation to enrol.
Book a Free Tech Assessment Find your nearest Robocode centre and book your child's Free Tech Assessment. Book a Free Tech Assessment
Does this change GCSEs or A levels?
No. This report is about universities and colleges in England. GCSEs and A levels are regulated separately.
Does my child need to know how to code first?
No. AI Story Studio is the first course on our AI pathway, for ages 11 to 17, and it starts from a blank file.
Will they just be using ChatGPT?
No. They build the app themselves and connect it to a real AI model. They write the prompt templates it uses and clean up what comes back.