During London Tech Week 2026 the UK government under Keir Starmer declared that it would provide 450,000 children on free school meals with AI tutors “to close the attainment gap”.
The plans have long been in the works. The Department for Education and the now-defunct Department for Science, Innovation and Technology released a press release in January that the government intended to use AI-powered learning tools to provide “personalised, one-to-one learning support – levelling the playing field for those who cannot afford private tutors”.
The benefits of private tuition to students are well-documented: Europe-wide studies show that 3 in 4 students who have online tutors report attainment improvements of up to three grades.
Nevertheless, the government’s announcement was met with swift backlash, with critics arguing that the program deepens educational inequality by providing poorer children with unproven chatbots whilst their wealthier counterparts continue to benefit from human tuition.
UK’s AI Tutoring Rollout Raises Questions Over Safety and Equality
Due to the relatively recent advent of widely accessible AI chatbots, their psychological impacts in a pedagogic environment remain largely unclear.
While some studies warn that frequent AI use could damage cognitive abilities relating to information processing and creativity, others show that AI-native learning environments can produce higher test scores and better academic engagement amongst students whilst also improving teaching methods and skills.
Concerns regarding AI in education, however, transcend the classroom: the UK’s NSPCC (National Society for the Prevention of Cruelty to Children) warns that AI can be used to extort, groom, harass, bully and mislead schoolchildren.
In fact, the UK government cited risks posed by the “intimate functionalities” of certain AI bots as one of the justifications for their social media ban for under 16s in June.
Danny Chambers, the Liberal Democrat MP for Winchester, spoke to 150Sec about his concerns about the policy.
“If AI can help children on free school meals achieve their potential and improve their life chances, that’s something we should explore”, the politician argued.
However, Chambers also warned that the government “must never create a system where wealthier children receive personalised human tuition while disadvantaged children are told that a chatbot is an adequate replacement. AI must never be used as a substitute for quality teaching or meaningful human interaction”.
The Starmer government had stated that its provision of virtual tutors is not intended to subvert or replace teachers.
“These [AI tools] should be tightly controlled educational tools rather than unrestricted chatbots. That means age-appropriate content, no addictive algorithms, no ability for strangers to contact children, and clear human oversight. These tools should undergo independent safety testing before being rolled out”, Chambers concluded.
The government announced in April that it would develop safe and effective AI tutoring tools for the recipients of free school meals under the supervision of teachers and in collaboration with EdTech and AI companies.
The companies selected for collaboration are Eedi, Eleven Labs, Learn Anything, Medly AI, Flint, Pearson (in partnership with Anthropic) and Zero Gravity. All partners will receive a contract worth £360,000, including VAT, for the partnership.
Google UK was also chosen as a partner, but has since withdrawn from the initiative.
AI and EdTech innovators back the policy, but warn that safeguards and purpose-built tools are necessary
Various leading figures in the EdTech field have spoken favourably about the policy.
David Pena del Rio, the CEO and founder of MiniTales, an interactive storytelling app that uses AI to generate personalized children’s books and audio stories, told 150sec that “Implemented responsibly and equitably, AI could become a powerful equalizer of educational opportunities for families of all kinds, regardless of their financial capacities”.
Ivan Crewkov, the CEO of purpose-built tutoring application Buddy.AI, which uses a voice-based AI tutor to teach English to young children and boasts over 76 million downloads worldwide, agrees.
He argued that AI is the only realistic way to make one-on-one tutoring, which he calls “the one proven method of dramatically improving learning outcomes”, affordable at scale.
However, Crewkov also spoke to 150sec about the need for the UK government to mitigate safety concerns in their rollout of AI tutors by foregrounding the protection of the privacy and security of children and their data.
Crewkov encouraged the British government to “test [the AI tools] in action and not just throw them out [to children] immediately” while ensuring “the [AI] tools are actually educational, they’re safe and that the privacy protection is provided”.
Crewkov’s Buddy.AI specifically pioneered its own AI infrastructure to ensure that the children’s voice recordings registered on his app are deleted and not routed through commercial APIs.
“They definitely need to work with proven and [child-protection] certified players on the market”, Crewkov continued.
Although many of the companies that the government is partnering with are built specifically for children, the government has admitted that the “AI-driven tools currently on the market are limited in quantity, scope and evidence base … with limited public access to the research and development that underpins them”,
The Incubator for Artificial Intelligence (i.AI), a unit of AI experts and civil service strategists working from within the UK government, is collaborating with the Department of Education on the nationwide implementation of the tools.
i.AI has announced that it will assess the efficacy of the AI tutoring tools according to their ability to “respond to typical educational interactions, including instructional support and safeguarding-relevant situations”.
The tools’ ability to integrate themselves into an educational setting and closely replicate the role of human tutors will be the key test for the government, according to Crewkov. They should not, he argues, be simple modifications of generative AI chatbots like ChatGPT.
“It is absolutely clear that those tools should be specifically built for teaching … and not optimized for pure engagement”.
Engagement with generative AI amongst British students is very high, with approximately 92% of them using generative AI tools to help with academic work. However, according to research by Oxford University Press, 1 in 4 British secondary school students believe that AI makes it too easy for them to find answers and complete assignments.
Sandra Falque, the founder of voice-led interactive storyteller for children Jules Stories, explained to 150sec how these future tutoring tools can help alleviate concerns associated with generative AI in schools.
“ChatGPT is here to answer your questions and … [therefore] there is a risk that people don’t make the effort to look for the answer themselves.”
Platforms like Jules, by contrast, can either be “trained to tell children the answer straight away or you can train Jules to ask the question in another way, like a teacher would do … guiding the child through the answer without saying the answer”, Falque explained.
“That is one example of how you can do what you want with this technology [AI] … one shouldn’t assume that it can only be a kind of ChatGPT”
That sort of methodology-centric approach, the founder argued, should be at the heart of the UK government’s program.
The AI tutoring trials are supposed to begin this autumn, while the final version of the tools is meant to be available to students by the end of 2027.
