Meta has disputed a report claiming that its Muse AI agent accessed a user’s private messages on a Mac without the user granting the required permission. According to the source, Meta says Muse cannot access Messages unless the user has explicitly enabled that access.

At this stage, the key point is that there is a disagreement between a published account and Meta’s explanation of how the product works. That means businesses should treat the underlying privacy and security question as unresolved unless further technical evidence or official clarification emerges.

What happened

A report alleged that Meta’s Muse AI agent read a journalist’s private messages even though the relevant Mac permission setting was turned off. Meta has rejected that claim, stating that Muse does not have access to a user’s Messages without explicit permission.

The source does not establish an independent technical finding, and the available information points to a dispute over what happened on the device and how the permission model operated in practice.

Why it matters for European businesses

For European companies, this is less about one disputed incident and more about a broader business issue: how AI assistants interact with local devices, business communications and personal data.

Many AI tools now promise to help users search files, summarise messages, draft replies and automate workflows across desktops and SaaS platforms. Those features can be useful, but they also raise questions about:

• what data an AI tool can access by default
• whether consent and permissions are clearly presented to users
• how local operating-system permissions interact with app-level settings
• whether employee or customer data could be exposed unintentionally
• how organisations document lawful use of AI tools that process personal data

In Europe, these issues connect directly to privacy governance, internal security controls and vendor risk management. Even when a vendor denies a specific allegation, the incident can still influence procurement decisions, internal AI policies and employee trust.

Who may be affected

The immediate relevance is highest for organisations that are testing or deploying AI assistants on employee devices, especially tools that can access messages, documents, calendars or browser content.

• SMEs and founders adopting AI copilots without a dedicated security review process
• IT teams responsible for device permissions, endpoint controls and approved software policies
• Marketing and e-commerce teams using AI tools that may process customer communications, campaign assets or internal planning documents
• Regulated businesses handling sensitive personal, financial, legal or health-related information
• Digital agencies working across multiple client accounts and datasets on shared workflows

What companies should consider

Businesses do not need to wait for the outcome of a public dispute to strengthen their AI governance. Practical steps include:

• Review AI assistant permissions on company-managed devices, including access to messages, files, contacts and screen content
• Limit use cases for AI tools where sensitive internal or customer data is involved unless access controls are well understood
• Check vendor documentation carefully to distinguish confirmed product behaviour from marketing claims
• Update internal AI policies so staff know which assistants are approved and what business data they may process
• Assess privacy and security implications before enabling integrations that connect AI tools to communications systems or local desktop content
• Keep auditability in mind by documenting permissions, approved workflows and data-handling decisions

For European businesses, the practical lesson is straightforward: AI productivity tools can offer real value, but they should be evaluated with the same caution applied to other software that may touch confidential communications or personal data.