Microsoft agrees to new AI safety rules for US schools
Microsoft Agrees to New AI Safety Rules for US Schools
Microsoft has reached a deal with the United States’ second‑largest teachers’ union to impose strict safety and privacy safeguards on its artificial‑intelligence tools used in K‑12 classrooms. The pact comes as educators and parents push back against the growing presence of AI and concerns over excessive screen time.
AI Safety Rules
The agreement bars any AI companion or feature that is designed to create emotional attachment, dependency, or to keep students engaged beyond the time needed for a specific learning activity. By eliminating such “sticky” functionalities, Microsoft aims to prevent AI from becoming a distraction or a source of undue influence in the classroom.
Data Use Restrictions
Under the terms, Microsoft will not employ student or teacher data to train its AI models, except for a narrow carve‑out that allows the use of information that could help protect children. The company also commits that the data it gathers cannot be sold, leveraged for advertising, or applied to the development of unrelated products. These provisions are intended to keep personal educational information confined to its original educational purpose.
Transparency Requirements
The deal includes audit and transparency obligations that require Microsoft to provide families with clear, plain‑language explanations of how its tools operate and what data they collect. Brad Smith, Microsoft’s vice chair and president, said the arrangement “sets a high bar for child privacy and AI safety” and that the company plans to roll the standards out to every school district nationwide. The new requirements will take effect for all Microsoft school contracts beginning November 1.
European Context
Across the Atlantic, AI is already finding its way into classrooms. A European Commission study of early adopters in Finland, Germany, Ireland, Luxembourg and Spain reported that students use generative AI to clarify difficult concepts, practice skills and prepare for exams. An Austrian study from 2026 showed that 60 percent of 11‑ to 17‑year‑olds had asked AI chatbots for advice on stress, conflicts or heartbreak, while 30 percent discussed personal worries with the bots. Norway recently advised that children aged six to twelve generally avoid using generative AI for schoolwork, allowing cautious experimentation for ages thirteen to fifteen. France permits primary‑school students to learn about AI but restricts direct use of generative tools, with supervised classroom applications permitted around age thirteen or fourteen. A document viewed by Euronews indicates the European Commission will propose a legislative ban on risky online services, including AI chatbots, for under‑15s.
Expert Concerns
Industry observers note that while the U.S. agreement introduces meaningful safety measures, it should not be taken as a green light for broader AI adoption in schools. Josh Golin, executive director of the online‑safety nonprofit Fairplay, warned that a high‑profile framework might be misread by districts as an endorsement, prompting them to deploy the tools without further evaluation. OpenAI and Anthropic are also in talks with the teachers’ union about similar safety and privacy pacts, but Google, the dominant education‑technology provider in American schools, has not disclosed whether it will offer comparable protections.
Source: euronews.com · 2026-09-15