Meta introduces muse, a secure AI agent with enhanced privacy features

Meta has launched Muse, a personal AI agent designed to automate digital tasks within a secure cloud environment, emphasizing privacy and security. Available on iOS, Android, and via the Muse.ai website, the tool also integrates with WhatsApp and Meta’s AI glasses. Muse is part of Meta Superintelligence Labs, established a year ago to compete with OpenAI and Anthropic. The agent can perform tasks like sending emails, booking travel, and managing purchases using Stripe’s Link payment infrastructure, which issues single-use card numbers to protect user financial data. Muse’s architecture, Secure VM, isolates user activity in a virtual machine to prevent untrusted data from interfering with actions. David Singleton, vice president of engineering for consumer products at Meta Superintelligence Lab, highlighted the Sentinel system, which checks actions against user policies or prompts human approval. Meta is also introducing Confidential VM, allowing users to manage access keys locally and ensuring no third party, including Meta, can access the VM. The company is offering a $300,000 bounty for vulnerability findings, including prompt injection attacks.