Microsoft Sets AI Guardrails for US Schools

Microsoft has introduced legally binding AI guardrails and privacy standards for its AI products in US schools, negotiated with the country’s second-largest teachers' union [1]. This move comes amid increasing scrutiny of AI use and screen time in education, setting a new benchmark for how tech companies manage student data and AI interactions.

New Standards for AI in Education

The agreement, signed last week, prohibits AI features "designed to foster emotional attachment or dependency" or to keep students engaged beyond a learning task [1]. These standards are legally binding and require third-party audits to ensure compliance [1]. Microsoft also committed to not using student or educator data to train AI systems, with a narrow exception for data that protects students, and will not sell this data, use it for advertising, or for product development [1]. The company will also provide transparent explanations to families about how its tools function and what data they collect [1]. Brad Smith, Microsoft's vice chair and president, stated that this agreement sets a high bar for child privacy and AI safety, and will be extended to every school district across the US [1]. The new standards will apply to Microsoft's school contracts nationwide starting November 1 [1].

Other major AI developers, OpenAI and Anthropic, are also in discussions with the American teachers’ union regarding similar safety and privacy pacts [1]. OpenAI, maker of ChatGPT, called Microsoft's agreement an "important milestone for AI safety and privacy in schools" and expects to finalize its own agreement [1]. Anthropic, which developed Claude, also supports responsible AI in education and is working with the union [1]. In contrast, Google, a dominant provider of education technology in US schools, has not indicated whether it plans to offer similar protections [1]. Google faced criticism after expanding its Gemini-powered tools in Google Classroom to school-issued accounts for students under 18 in August [1]. Google states that Classroom is used by over 150 million teachers and students globally [1].

Broader Regulatory Landscape

The push for AI guardrails extends beyond the US. In Europe, AI is already present in classrooms, with a European Commission study finding students in Finland, Germany, Ireland, Luxembourg, and Spain using generative AI for explanations, practice, and exam preparation [1]. An Austrian study from 2026 found that 60% of 11- to 17-year-olds surveyed sought advice from AI chatbots for issues like stress or conflicts, and 30% discussed worries or feelings with them [1]. Some European governments are already setting limits: Norway advises children aged 6-12 generally avoid generative AI for schoolwork, while those 13-15 can experiment cautiously [1]. France allows primary school pupils to learn about AI but not directly use generative AI, with supervised classroom use starting around ages 13-14 [1]. A legislative proposal from the European Commission is expected to propose barring under-15s from risky online services like AI chatbots [1].

What to watch

• Competitor responses: Monitor whether Google, a major education technology provider, will adopt similar AI safety and privacy standards for its tools in schools [1].
• Union negotiations: Watch for the finalization of agreements between the American teachers' union and other AI companies like OpenAI and Anthropic regarding their own AI safety and privacy pacts [1].
• European regulation: Observe the progress of the European Commission's legislative proposal regarding online services and AI chatbots for users under 15 [1].

The Microsoft agreement establishes a significant precedent for AI governance in educational settings, potentially influencing broader industry standards for child privacy and data use. The effectiveness of these guardrails will depend on their widespread adoption across the tech sector and consistent third-party oversight.