Listen to this story Members can listen to an AI-generated audio version of this article. 1.0x Audio narration uses an AI-generated voice. 0:00 0:00 Become a member to listen to this article Subscribe The government has contracted the recruitment agency Randstad to find hundreds of teachers to test new safety standards of AI tutoring tools. The government’s AI incubator project is creating “sovereign education benchmarks” to evaluate whether AI tutors deliver effective, safe and curriculum-aligned teaching. It will be working with teachers to create example classroom interactions and clear scoring criteria for responses. The Department for Science, Innovation and Technology (DSIT) said the hands-on input would help to ensure benchmarks were “accurate, aligned to teaching standards and safe to use”. It said that rather than relying on “market claims or general-purpose performance scores” benchmarks evaluated AI systems using “realistic scenarios, clear success criteria and consistent scoring methods”. The DSIT will recruit teachers for this work through Randstad, under a two-month contract worth £900,000. No further details were provided on how they would be recruited, or what criteria they would need to meet. Randstad is one of the largest HR consulting and staffing companies in the world. It was famously axed as the contractor for the Covid catch-up national tutoring programme in 2022 after a tumultuous year running the scheme. Curriculum support A DSIT spokesperson said: “AI tutoring tools are being co-designed with teachers to give pupils access to the kind of curriculum support usually only available to families who can afford private tutoring. “We’re working with hundreds of teachers to build benchmarks and scoring criteria, so these tools meet strict safety standards before they reach classrooms. “We’re recruiting teachers for this work through Randstad.” The government hopes AI tutoring tools will help level the playing field for pupils who cannot afford private tutors. But Becky Francis, the chief executive of the Education Endowment Foundation (EEF), has previously stressed the evidence “on AI provision is in its infancy”. Susan Acland-Hood, the Department for Education’s permanent secretary, told MPs last month they had seen “quite promising evidence that both good-quality AI tutors and good-quality AI support for tutoring that integrates the work that humans are doing can have what look like positive results”. “But what we are doing is testing people who have some promising products, and we will evaluate that incredibly carefully and thoughtfully, and look at what works. “If it does not work, we will not do it, because that would be silly.”