Meta launches AI agent that can access other apps to send emails, make payments - Reuters

Meta has introduced a new artificial‑intelligence agent that can interact with external applications on behalf of users. The agent is designed to operate across a range of services, extending the functionality of Meta’s existing AI offerings.

According to the announcement, the agent is capable of accessing other apps to perform tasks such as composing and sending emails, as well as initiating payments. These actions are carried out through the agent’s ability to interface directly with third‑party platforms, allowing users to complete routine activities without leaving the Meta environment.

The rollout is part of Meta’s broader effort to embed generative AI tools throughout its product suite. The company has positioned the agent as a means to streamline everyday digital interactions, leveraging its AI models to understand user requests and execute them across connected services.

Internal reports have highlighted concerns that the technology could mismanage access to sensitive personal data. Employees have flagged the risk that the agent’s broad permissions might lead to unintended exposure or misuse of private information when interfacing with external applications.

Meta has acknowledged the internal feedback and indicated that safeguards are being evaluated to address data‑privacy risks. The company’s statements suggest ongoing work to ensure that the agent’s access controls and data handling practices meet its security standards while the feature is further developed.