Order bakery pickup and find couches on Marketplace
Published September 20, the review described useful browsing and shopping assistance alongside repeated prompts to connect sensitive accounts and personal information. Task: Published September 20, the review described useful browsing and shopping assistance alongside repeated prompts to connect sensitive accounts and personal information. Key points 01 A hands-on review published September 20 described one several-day test of Meta Muse. A hands-on review published September 20 described one several-day test of Meta Muse. 02 In that test, Muse added a sandwich to a bakery pickup cart and found couches on Facebook Marketplace; checkout paused for final approval. In that test, Muse added a sandwich to a bakery pickup cart and found couches on Facebook Marketplace; checkout paused for final approval. 03 The test observed prompts to connect financial accounts, email, documents, and passport or license expiration details. The test observed prompts to connect financial accounts, email, documents, and passport or license expiration details. 04 Meta says the setting for using Muse interactions to improve AI models is on at first use and can be turned off in Data controls. Meta says the setting for using Muse interactions to improve AI models is on at first use and can be turned off in Data controls. A September 20 hands-on review found useful task assistance and repeated prompts for sensitive data. Meta’s training and memory controls work differently. A hands-on review published September 20 found that Meta Muse could help with web tasks while repeatedly suggesting that the user connect sensitive information. In the several-day test, Muse added a bakery order to a pickup cart and searched for couches on Facebook Marketplace; it also prompted the user to connect financial accounts, email, documents, and identity information. Meta announced Muse on September 8, 2026, describing it as a personal AI agent and announcing an initial U.S. rollout through iOS, Android, and the web. What Muse did in the hands-on test Muse added a bacon, egg, cheddar, and garlic-aioli biscuit sandwich to a Kahnfections pickup cart. The checkout process paused for final approval. In a separate task, Muse found local couches on Facebook Marketplace and followed up about a listing the user liked. Those examples show the appeal of delegating steps to an agent: it can browse and prepare an action, rather than just suggest what you could do. In the test, the user still had to approve the checkout. The prompts reached into sensitive categories The test recorded repeated suggestions to connect checking and savings accounts, scan an inbox, provide documents, and share passport or driver’s license expiration details. The prompts were observed; they do not mean every account or document mentioned was connected. That distinction matters when an agent’s usefulness depends on access to personal context. A suggested connection is not the same as a completed connection, but it can still shape what information a user considers sharing. Meta’s controls have different jobs Meta says each user’s Muse runs in an isolated cloud virtual machine, or VM. The company describes Sentinel as a separate system that governs permissions for connected services and outbound network activity. It also says credentials are stored apart from the agent’s runtime, so the main agent does not see real credentials. Meta’s description of the launch architecture also says the company can access VM data when needed to operate, support, or secure the service. Isolation and permission checks are part of Meta’s stated design; they do not prevent that operational access. Meta says Muse conversations and VM data are not shared with its advertising systems. Browsing or shopping through Muse can still affect ads indirectly through ordinary web-tracking signals. Training data and Memory are separate controls Meta says the setting that allows Muse interactions to be used to develop and improve AI models is on when someone first uses Muse. Users can turn it off under Data controls, and Meta says the change also applies to previous interactions. Memory is a different feature: it retains information and preferences for future tasks. In the tested version, the Memory document could be edited or cleared by asking Muse, but the reviewer found no switch to disable Memory altogether. Meta’s help guidance says deleted content may remain in learned memories until the user uses Muse’s forget function. Other Languages Read this article in another available language. NeoTeo Read next Chromebook vs. Googlebook: Which Fits Your Workflow?; tools: Meta Muse. Human role: The original post does not say how a person stayed involved.; limitations: Muse runs in an isolated cloud VM, but Meta retains operational access to VM data for support and security. The training data setting is on by default and must be manually disabled in Data controls. Memory features cannot be fully disabled, only edited or cleared, and deleted content may persist in learned memories. Users must approve all checkout actions, and the agent's prompts for sensitive data may raise privacy concerns..
Tools
Published on Agent Work Atlas:
Overview
Published September 20, the review described useful browsing and shopping assistance alongside repeated prompts to connect sensitive accounts and personal information.
Workflow
Published September 20, the review described useful browsing and shopping assistance alongside repeated prompts to connect sensitive accounts and personal information. Key points 01 A hands-on review published September 20 described one several-day test of Meta Muse. A hands-on review published September 20 described one several-day test of Meta Muse. 02 In that test, Muse added a sandwich to a bakery pickup cart and found couches on Facebook Marketplace; checkout paused for final approval. In that test, Muse added a sandwich to a bakery pickup cart and found couches on Facebook Marketplace; checkout paused for final approval. 03 The test observed prompts to connect financial accounts, email, documents, and passport or license expiration details. The test observed prompts to connect financial accounts, email, documents, and passport or license expiration details. 04 Meta says the setting for using Muse interactions to improve AI models is on at first use and can be turned off in Data controls. Meta says the setting for using Muse interactions to improve AI models is on at first use and can be turned off in Data controls. A September 20 hands-on review found useful task assistance and repeated prompts for sensitive data. Meta’s training and memory controls work differently. A hands-on review published September 20 found that Meta Muse could help with web tasks while repeatedly suggesting that the user connect sensitive information. In the several-day test, Muse added a bakery order to a pickup cart and searched for couches on Facebook Marketplace; it also prompted the user to connect financial accounts, email, documents, and identity information. Meta announced Muse on September 8, 2026, describing it as a personal AI agent and announcing an initial U.S. rollout through iOS, Android, and the web. What Muse did in the hands-on test Muse added a bacon, egg, cheddar, and garlic-aioli biscuit sandwich to a Kahnfections pickup cart. The checkout process paused for final approval. In a separate task, Muse found local couches on Facebook Marketplace and followed up about a listing the user liked. Those examples show the appeal of delegating steps to an agent: it can browse and prepare an action, rather than just suggest what you could do. In the test, the user still had to approve the checkout. The prompts reached into sensitive categories The test recorded repeated suggestions to connect checking and savings accounts, scan an inbox, provide documents, and share passport or driver’s license expiration details. The prompts were observed; they do not mean every account or document mentioned was connected. That distinction matters when an agent’s usefulness depends on access to personal context. A suggested connection is not the same as a completed connection, but it can still shape what information a user considers sharing. Meta’s controls have different jobs Meta says each user’s Muse runs in an isolated cloud virtual machine, or VM. The company describes Sentinel as a separate system that governs permissions for connected services and outbound network activity. It also says credentials are stored apart from the agent’s runtime, so the main agent does not see real credentials. Meta’s description of the launch architecture also says the company can access VM data when needed to operate, support, or secure the service. Isolation and permission checks are part of Meta’s stated design; they do not prevent that operational access. Meta says Muse conversations and VM data are not shared with its advertising systems. Browsing or shopping through Muse can still affect ads indirectly through ordinary web-tracking signals. Training data and Memory are separate controls Meta says the setting that allows Muse interactions to be used to develop and improve AI models is on when someone first uses Muse. Users can turn it off under Data controls, and Meta says the change also applies to previous interactions. Memory is a different feature: it retains information and preferences for future tasks. In the tested version, the Memory document could be edited or cleared by asking Muse, but the reviewer found no switch to disable Memory altogether. Meta’s help guidance says deleted content may remain in learned memories until the user uses Muse’s forget function. Other Languages Read this article in another available language. NeoTeo Read next Chromebook vs. Googlebook: Which Fits Your Workflow?
Reported result
The review confirmed that Muse could successfully prepare shopping carts and find listings, but it did not report measured business outcomes or efficiency gains. The agent's utility was limited by its tendency to request access to sensitive personal information, which the reviewer observed but did not necessarily grant. No specific metrics on time saved or cost reduction were provided in the source material.
Human role
The original post does not say how a person stayed involved.
Limitations
Muse runs in an isolated cloud VM, but Meta retains operational access to VM data for support and security. The training data setting is on by default and must be manually disabled in Data controls. Memory features cannot be fully disabled, only edited or cleared, and deleted content may persist in learned memories. Users must approve all checkout actions, and the agent's prompts for sensitive data may raise privacy concerns.
Try it
I want to test Meta Muse for personal shopping assistance. Please guide me through a dry run where I ask Muse to add [replaceable item, e.g., a specific sandwich] to a [replaceable store, e.g., Kahnfections] pickup cart and search for [replaceable item, e.g., couches] on Facebook Marketplace. Before executing any checkout or connecting any accounts, pause and show me the proposed action for my approval. Do not connect any financial accounts, email, or identity documents unless I explicitly instruct you to do so. After the dry run, check the Data controls to ensure the 'Use interactions to improve AI models' setting is turned off. Report back on whether the agent successfully prepared the cart and listing without requiring sensitive data access.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.