Plans for written coursework will face “far, far more scrutiny” under reformed qualifications to prevent the “AI-fuelled subversion of assessment”, the boss of England’s assessment regulator will say. Ofqual boss Ian Bauckham is set to issue the warning at the Festival of Education today ahead of the planned launch of the first three V Level subjects next year. “For the qualifications to be worth more than the paper they are written on, we cannot normalise the idea that AI-generated output is a substitute for genuine human endeavour,” he will say. ‘Subversion’ Bauckham will point to examples from higher education of universities struggling to grapple with the issue, including the UCL Law School decision to overhaul assessments to make them “AI-proof by making greater use of in-person formats”. He will argue that unis “are fast waking up to the harsh reality that if they want their qualifications to continue to be taken seriously”, then they must get a handle on “AI-fuelled subversion of assessment”. “So as we design new qualifications for 16- to 18-year-olds, and reform existing ones, expect to see far, far more scrutiny of any proposals for written coursework. “For the qualifications to be worth more than the paper they are written on, we cannot normalise the idea that AI-generated output is a substitute for genuine human endeavour.” Bauckham will also caution against the wider use of the tech in the classroom, arguing it prompts “less learning, less satisfaction, less usefulness in the workplace, and ultimately is a covert assault on the realisation of human potential”. AI marking But on the use of AI in the qualification system, he acknowledged it “has enormous potential to improve efficiency and can both cut costs and increase accuracy. It is doing so already.” However, it is “not actually as good as you might think” at marking, “despite what the tech-enthusiasts would have you believe”. This is because it “makes mistakes, it hallucinates, and it reflects biases”. It can also “extend beyond the rules initially set by humans, particularly when dealing with cases not anticipated in advance”. “AI will have an important place, but we must remain in charge,” he will add. “We must legislate and regulate to create clarity about what AI can be allowed to do to bring about benefits for humans, and what it absolutely cannot.”
Clare Russell 7 July 2026 This is an important debate, but I wonder if we’re framing the question too narrowly. Rather than asking how we protect existing assessments from AI, perhaps we should be asking how assessment itself needs to evolve. AI can increasingly generate content, but it can’t replace human judgement, ethical decision-making, creativity or the ability to apply knowledge in complex situations. Rather than asking students simply to write an essay, we could ask them to develop an argument using AI as a research partner, explaining why they accepted some evidence and rejected other sources. Business students could use AI to analyse a market before presenting and defending their own strategic recommendations. Engineering students might use AI to generate design options, then justify the solution they select against cost, safety and sustainability criteria. Health and social care students could use AI to explore treatment options before explaining how they would adapt their approach to meet an individual’s needs. Those are the kinds of capabilities employers increasingly value. We are facing a once-in-a-generation opportunity to redesign assessment so that it rewards the distinctly human skills of reasoning, judgement, creativity and professional decision-making, rather than simply producing content that AI can now generate. Let’s not pass that up!