Meta is introducing privacy features to its AI services, acknowledging user concerns about data usage. CEO Mark Zuckerberg announced a commitment to a fully private mode for personal agents, ensuring user information remains inaccessible to Meta or other providers. The Muse AI assistant allows users to control app connections and opt out of data training for models. Interactions with Muse will not be used for Meta's ad systems, and a more secure confidential VM option is planned. Meta also intends to implement private cloud processing for its AI glasses by year-end. The author has tentatively begun using Muse, disabling its data training capabilities and awaiting further privacy upgrades. Current uses include receiving notifications about sports schedules and getting product recommendations. The author previously found Meta's AI glasses appealing for capturing moments but was deterred by data usage for ad targeting. However, this endorsement is tentative due to Meta's past privacy issues and unaddressed bystander privacy concerns. Electronic Privacy Information Center warns of serious privacy risks associated with Meta's camera and microphone-equipped devices. The author is willing to try Meta's AI offerings but remains cautious about fully trusting the company with sensitive data.