Skip to content

Mews shifts Meta from conversation to execution as an independent task agent tests trust levels via WhatsApp and the browser

Share
Mews shifts Meta from conversation to execution as an independent task agent tests trust levels via WhatsApp and the browser

Listen to this article

Read by Anchor

Meta announced the launch of its new smart agent, Mews, for users in the United States, in a step that represents an architectural shift from the generation of traditional chatbots that answer questions and hold conversations to a stage of agents capable of taking on tasks and executing them independently in daily consumer life. The announcement, which came less than two weeks after the company’s approval of a major $18 billion legal settlement covering multiple U.S. states over social-media harms, puts the company before a precise trust test, as operating the agent requires granting Meta deeper access to personal data and workflow than any previous communication platform.

The agent relies on the “Mews Spark” model developed by the company and connects directly with the applications and services an individual uses regularly, such as email, calendar, payment platforms, smart-home management, bookings, fitness, and shopping. The system’s role is not limited to offering suggestions; it also sends emails, books travel, reduces bills, fills out forms, organizes event lists, converts video cooking clips into grocery shopping lists, and completes purchases directly using Stripe’s Link service, which provides purchase protection to ease consumer concerns, while the company prepares for later integration with Shopify’s Shop Pay service and the password manager 1Password.

Mews goes beyond closed conversational interfaces to operate in the background and learn from user conversations, offering proactive suggestions even after the app is closed.Meta designed the system to work through built-in software connectors that support public APIs using credentials supplied by the user, or through autonomous web browsing when an API is unavailable. The agent is initially available via the website, iOS and Android mobile apps, and WhatsApp messages, with plans to embed it in Meta’s smart glasses soon. Although the service starts free, registration requires a payment card; paid plans activate as usage increases, offering two options: the Power tier at $20 per month and the Maximum tier at $100 per month, with a consumption meter that tracks the remaining quota and alerts the user when the free allotment is exhausted.

On the security side, the company claims to run the agent inside a dedicated secure virtual machine equipped with its own browser called Mews Secure VM, alongside an independent monitoring agent named Sentinel that operates in the same virtual environment but is isolated at the system level to prevent Mews from seeing passwords or payment methods, and the company emphasizes that data or conversations will not be shared with its advertising system. However, these technical safeguards face a long legal legacy, starting with the Federal Trade Commission settlement in 2011, a $5 billion fine in 2019 for privacy violations, a 2023 accusation of violating an executive order, the Cambridge Analytica scandal and the 2019 leak of unencrypted passwords for millions of accounts, and recent cases that included a $942 million judgment in New Mexico for protecting minors.

In our assessment, moving executive agents to WhatsApp represents the most important variable for the business and technology environment in the Gulf region, Egypt, and the Arab Levant.In these markets, where the app serves as the primary channel for customer service, e-commerce, and communication between businesses and consumers, the introduction of an agent that executes payments and manages forms on behalf of the individual represents a fundamental shift in commercial interaction rules; digital presence will no longer be limited to bots that answer inquiries, and commerce platforms, banks, and software developers will need to equip their APIs to receive automated purchase requests executed by a digital agent with human authorization. This trajectory places regional tech and compliance teams in a position where they must scrutinize data-protection policies, especially since delegating an automated agent to handle daily transactions requires precise alignment with national personal data laws in countries such as Saudi Arabia, the United Arab Emirates, and Egypt.

Meta’s entry into the independent-agent pathway alongside other market experiments such as Gemini Spark and Cloud Co-Work confirms that competition has moved beyond text generation and has fully shifted to execution engineering and task management. Although the company has allowed consumers to customize the agent’s appearance, avatar, name, and communication style, the real bet remains tied to the resilience of the security-isolation architecture and to how willing users and organizations are to grant financial and procedural authorizations to a system that is subject to ongoing regulatory oversight.

Don't miss the next story

Subscribe for updates