Both labs claim their agents guard user data better than the competition, but Meta’s Muse faced a zero-day flaw and late security issues.

OpenAI used its DevDay keynote to introduce Dots, a new AI agent that CEO Sam Altman said would set a new standard for privacy in frontier AI. OpenAI spent the day contrasting Dots with Meta’s Muse agent, arguing that its rival fails to keep user data safe. Meta launched Muse months earlier as a safer successor to OpenClaw.

Meta CEO Mark Zuckerberg described Muse as built from the ground up for privacy and security, with user data stored on an isolated Linux virtual machine. Nat Friedman, head of product at Meta Superintelligence Labs, said the company aimed to make something like OpenClaw that it could keep safe and scale to billions. Meta warned Muse can still make mistakes.

Enterprise teams evaluating agents should treat vendor privacy claims as marketing until they verify them. Meta says it can still access data inside a customer’s VM, even though that data stays isolated from other users. A security researcher found a zero-day flaw that allowed takeover of Muse before Meta patched it, and 404 Media reported last-minute issues that could have exposed internal Meta databases.

Meta plans to introduce a method that cryptographically and verifiably prevents Meta from accessing data in a user’s VM later this year. Until independent researchers test that claim, operators should assume the company retains access. Agent vendors will keep trading privacy promises as a differentiator, so buyers should ask for verifiable controls rather than marketing language.

What matters

• OpenAI unveiled Dots at DevDay and promised a new privacy standard for frontier AI agents.
• Builders adopting agents inherit data-access risk that vendor privacy pledges do not remove.
• Meta plans cryptographic proof that it cannot read VM data, and that claim deserves independent testing.

Why it matters

Meta plans cryptographic proof that it cannot read VM data, and that claim deserves independent testing.

This GenAI News article was prepared in original wording using reporting and materials published by The Verge. Source reference: https://www.theverge.com/ai-artificial-intelligence/1009051/privacy-ai-agent-promises-openai-meta-muse-dots.

Drafted by the GenAI News review pipeline.