Meta has denied allegations that its Muse AI agent accessed a user’s private messages content without permission, arguing that the feature requires several explicit permissions on a Mac before it can function.
The dispute follows a report by journalist Jason Aten, who said Muse accessed his messages despite Full Disk Access being disabled.
According to TechCrunch, Aten said he asked the AI how it obtained the content and was told it was syncing device notifications, leading him to suspect that incoming message previews were being sent to the agent.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
Meta vice president of communications Andy Stone rejected that explanation. He said the Messages connection in Muse is optional and cannot access message content unless users activate both Full Disk Access and the Messages connector.
Meta executive David Singleton provided a more detailed explanation, saying users must complete multiple permission steps involving Muse, macOS settings and access to Messages.
Also Read | Facebook Marketplace Mishap: Meta User Says Muse AI Gave Out His Address, Agreed To Pickup He Never Approved
Users can choose no access, read-only access or read access after enabling Full Disk Access. Singleton said these protections cannot be bypassed even if the application contains a bug.
He also disputed Muse’s explanation about syncing notifications, saying the AI had provided an inaccurate account of what happened. Meta pointed to Muse’s security architecture and bug bounty programme in its response.
The disagreement has raised fresh questions about how users can verify what AI agents are accessing on their devices. TechCrunch noted that trust will be a key issue for Muse as Meta seeks a larger position in the consumer AI market.
The controversy comes as another Muse-related incident draws attention. YouTuber Matt Robb said the AI mishandled a Facebook Marketplace task, resulting in his address being exposed and a prospective buyer arriving while he was away. Singleton has reportedly indicated that Meta is examining that case.
Meta’s response to Aten is clear that the company says the reported access should not have been technically possible under its permission model.
Also Read | Meet Muse: Meta’s New AI Agent For ‘Real-World’ Tasks – Features And How It Works
However, the conflicting accounts leave open questions about what occurred on the device and whether a technical issue was involved.
Essential Business Intelligence,
Sharp Market Insights,
Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
