Questions raised about the security of the new AI agent 'Muse: Find out just how safe Meta's new tool is..

Imagine needing to book a flight, order groceries online, speak to customer service, or schedule an appointment—without having to do any of the work yourself. You simply tell an AI agent what needs to be done, and it completes the task on your behalf.

Meta is preparing for a future just like this. The company wants users to utilize its AI agent, Muse, for everyday tasks. After being unveiled by CEO Mark Zuckerberg at a recent conference, Meta’s new Muse quickly climbed to the top of app charts.

Muse is designed so that users do not require any special training to operate it. It can order groceries online, call customer service, send emails, book appointments, and plan travel itineraries for the user.

However, alongside the convenience offered by the growing capabilities of AI agents, serious questions are also arising. The biggest concern is: if people begin delegating more and more of their daily tasks to these agents, what impact will it have on their privacy and data?

What are AI agents?
First, it is important to understand exactly what they are. AI agents are systems that employ technologies such as Large Language Models (LLMs) and reinforcement learning to achieve goals set by the user.

While traditional chatbots generally answer questions, AI agents go a step further; they can independently take action across multiple stages to complete a specific task.

The capabilities of AI agents have evolved rapidly over the past 12 months. Their use has expanded across a wide range of applications, from DIY hobby projects to advanced personal assistant apps. Amidst this, there have been reports of high-profile hacks and system breaches involving "rogue" agents—uncontrolled AI agents. Consequently, Meta's entry into this space is seen as a significant step toward bringing AI agents to the general public.

**Meta’s Major Investment in the AI ​​Race**

Meta is striving to solidify its position in the rapidly evolving AI landscape. In 2025, the company spent $14.3 billion to bring Scale AI entrepreneur Alexandr Wang onto its team and subsequently completely revamped its internal AI systems.

In total, Meta spent $72 billion on AI and other capital expenditures in 2025. The company is expected to double this spending in 2026. However, despite these massive investments, Meta's AI models do not currently lead the pack globally.

**Meta’s AI Model Ranks Third**

In terms of rankings based on popularity, Meta’s latest AI model, Muse Spark 1.2, ranks third—trailing behind the top-tier Claude and GPT models from Anthropic and OpenAI. Yet, this may not be Meta's biggest challenge. Anthropic and OpenAI primarily target professionals and business users willing to pay for the most capable models for tasks such as coding, research, and analysis.

For instance, OpenAI’s recently launched "Dots" agent targets customers seeking AI assistance for business-related tasks. Meta also plans to offer professional-grade paid AI services, but its greatest strength lies in its massive user base. Meta Wants Accessible AI, Not Just the Smartest
In the June quarter, 3.6 billion people used Facebook, Instagram, Messenger, or WhatsApp daily. These users do not necessarily require the world's smartest AI model; instead, Meta is focusing on AI that is easy to use and can reach the widest possible audience.

For this reason, the goal for the Muse Spark model isn't simply to be the most powerful model available. It needs to be capable enough to seamlessly power the Muse AI agent. Meta is promoting Muse as an AI agent designed for everyone—including those with little to no technical experience.

Muse to Soon Feature a Human-Like Avatar
Mark Zuckerberg has revealed that users will soon be able to create a friendly, human-like avatar for their Muse agent. Users will also be able to engage in real-time video chats with it. This means the AI ​​agent will evolve beyond a tool that operates solely via text or commands; it will feature a more personal, human-like interface. Zuckerberg also introduced a keychain-sized device that allows users to interact with the agent without needing a phone.

From Scrolling to Delegating Tasks
For the past two decades, Meta’s business model has relied on people's online habits. Notifications, endless feeds, and the validation received from others have created a habit of frequently checking Facebook and Instagram. Muse aims to steer this behavior in a new direction: instead of scrolling through social media feeds, users can delegate tasks to their AI agent.

Muse has been designed with this in mind. Users can name their agent and customize its appearance and speaking style. With every successfully completed task, the user's trust in the agent is likely to grow. Through the settings, users can also determine how many times the agent must seek their permission before completing a task.

Muse's reach is currently limited
For now, Muse's reach is limited, and it is available only in the US. Using connectors, the AI ​​agent can be linked to a user's email, calendar, and certain consumer services—such as restaurant bookings, music, and shopping. However, Muse cannot currently log in to a user's bank, medical, or tax accounts.

Meta plans to add more connectors over time and introduce the product to international markets.

Disclaimer: This content has been sourced and edited from Amar Ujala. While we modified it for clarity and presentation, the original content belongs to its respective authors and website. We do not claim ownership of the content.