2026 Responsible AI Transparency Report

Key Findings

• — Microsoft supported more than 450 Sensitive Use reviews and nearly 2,500 generative-AI product and feature pre-release reviews between July 2025 and June 2026; the Sensitive Uses team has processed over 1,900 submissions since 2019.
• — The new Sensitive Interactions Policy governs open-ended emotional interactions with AI systems and 'requires techniques to prevent anthropomorphic behavior and sets boundaries to help users avoid emotional entanglement, with enhanced safeguards for youth'; consumer mitigations include system responses to harmful content and unsafe behaviour including self-harm or suicidal ideation.
• — Analysis of Copilot usage found health-related queries dominate mobile usage and nearly one in five health-related conversations involve personal symptom assessment; Copilot Health went through a Sensitive Use review that flagged ungrounded health answers and conversations lacking context as risks.
• — Microsoft signed a memorandum of understanding with the Australian AI Safety Institute on human-AI interaction risks including overreliance and emotional dependency on AI companion chatbots, sits on the MLCommons AILuminate Psychosocial Task Force (crisis prevention, unqualified AI advice), and lists 'AI companionship' as a risk domain on its internal benchmark platform.
• — Youth input came through an AI Futures Youth Council and Cyberlite-facilitated co-design workshops producing guidance on AI companions written by teens for teens.

Methodology Notes

Corporate self-report with aggregated, de-identified usage analysis and no external audit; 35-page PDF created 29 August 2026 and released with a Microsoft On the Issues blog post on 1 September 2026. Reporting period 1 July 2025 to 30 June 2026. The Copilot health-usage figure cites Microsoft's separately published Copilot health-queries study (held).

Sources