Get the weekly research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean. No account.
The statement authorizes Microsoft to use personal data, including data generated through product use, to develop, train, and fine-tune AI models including large language models. In some markets, users can opt out of their conversation data being used to train AI models in Microsoft Copilot.
This analysis describes what Microsoft's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision establishes that personal data collected through Microsoft products may be used for AI model development and training, including LLMs. The opt-out mechanism is described as market-dependent, meaning it may not be available to all users, and the statement does not specify which markets have access to the opt-out.
Interpretive note: The statement does not specify which markets have access to the Copilot AI training opt-out, nor does it clarify the scope of data categories subject to AI training use across all Microsoft products.
The updated privacy statement removes the previous detailed list of third-party sources from which Microsoft obtains personal data, including data brokers, public social media posts, location service providers, co-branded partners, and developers. Under the revised language, Microsoft describes obtaining data from 'Microsoft affiliates, subsidiaries, and third parties' without specifying the categories or types of third parties as explicitly as before. The company states it has reorganized the document for greater clarity and accessibility, but the operational effect is that users receive less specific disclosure about where their data originates from outside Microsoft.
View change record →The updated policy establishes additional grounds on which Microsoft may retain personal data. While the prior version tied retention to specific user expectations and available deletion controls, the revised language authorizes retention for 'operating our business, meeting our contractual and legal obligations, improving and developing our products and services, protecting the safety and security of our systems and customers, and resolving disputes.' This expands the stated purposes beyond transaction fulfillment and legal compliance. The updated policy directs users to product-specific documentation for retention details rather than providing explicit deletion procedures and timelines in the privacy statement itself.
View change record →The updated policy now grounds data retention in five broad business purposes: operating the business, meeting contractual and legal obligations, improving and developing products and services, protecting system and customer safety, and resolving disputes. Previously, the policy articulated specific criteria for determining retention periods, including customer expectations for retention until manual deletion, availability of automated deletion controls, and data sensitivity. The revised language removes these granular criteria and instead requires users to consult individual product documentation to understand when their specific data will be deleted. This shifts the burden of finding retention timelines from the main policy statement to separate product-specific documents.
View change record →Under this provision, personal data generated through use of Microsoft products may be used to train AI models including LLMs. The agreement states that in some markets, users can opt out of conversation data being used for AI model training in Copilot, but does not confirm opt-out availability across all product categories or geographies.
Cross-platform context
See how other platforms handle AI Model Training from User Data and similar clauses.
Compare across platforms →Monitoring
Microsoft has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 25 platforms.
"As part of our efforts to improve and develop our products, we may use your data to develop, train, and fine-tune our AI models, including large language models (LLMs). Learn more here. We use data to develop new products. For example, we use data, often de-identified, to better understand our customers' computing and productivity needs, and to train and fine-tune AI models, which can shape the development of new products.Excerpt from Microsoft's Privacy Statement (Legacy)
(1) REGULATORY LANDSCAPE: This provision implicates GDPR Article 6 (lawful basis for processing) and Article 5 (purpose limitation), as using personal data originally collected for product delivery as training data for AI models may require a separate lawful basis or compatibility assessment under EU data protection law. The EU AI Act may require evaluation where LLM training constitutes development of a general-purpose AI model. Enforcement authorities include EU Data Protection Authorities and the UK ICO. The FTC has jurisdiction over data use practices that may constitute unfair or deceptive practices under Section 5 of the FTC Act. (2) GOVERNANCE EXPOSURE: High. The use of personal data for AI model training, particularly where the opt-out is described as available only in some markets, creates potential GDPR compliance exposure regarding lawful basis and purpose limitation. EU and UK data protection authorities have scrutinized AI training data use in enforcement actions against other technology companies, and the legal basis for this processing where consent is not obtained remains contested. (3) JURISDICTION FLAGS: EU/EEA and UK users face the highest regulatory exposure given GDPR purpose limitation and lawful basis requirements. California users may have rights under CCPA to opt out of uses of personal information that qualify as sale or sharing. Users in markets without an explicit opt-out mechanism have limited recourse under the terms as written. (4) CONTRACT AND VENDOR IMPLICATIONS: Organizations procuring Microsoft enterprise products should verify whether their organizational data processed under the Product Terms and Data Protection Addendum is excluded from this AI training use. The statement indicates enterprise products are governed by separate agreements; however, consumer-facing Copilot products used with personal Microsoft accounts are subject to this provision. Procurement teams should confirm contractual carve-outs for enterprise data. (5) COMPLIANCE CONSIDERATIONS: Compliance teams should assess whether current consent or notice mechanisms adequately disclose AI training use to users at the point of data collection, particularly for Copilot and Microsoft 365 consumer subscribers. Data mapping exercises should document which product data flows may be used for AI training purposes. Organizations with EU or UK users should evaluate whether a legitimate interests assessment or alternative lawful basis documentation is required for this processing.
Full institutional analysis
Regulatory citations, enforcement risk, and due diligence action items.
Monitor: same-day alerts on the platforms you choose. Analyst: full institutional analysis.
Compliance Governance Intelligence
Need to monitor specific governance provisions?
Compliance includes provision-level monitoring, governance timelines, regulatory mapping, and audit-ready analysis.
Built from archived source documents, structured governance mappings, and historical version tracking.
This provision establishes that personal data collected through Microsoft products may be used for AI model development and training, including LLMs. The opt-out mechanism is described as market-dependent, meaning it may not be available to all users, and the statement does not specify which markets have access to the opt-out.
Under this provision, personal data generated through use of Microsoft products may be used to train AI models including LLMs. The agreement states that in some markets, users can opt out of conversation data being used for AI model training in Copilot, but does not confirm opt-out availability across all product categories or geographies.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Microsoft.