Meta’s Muse AI Agent Struggles With Basic Travel Booking Tasks
An independent test reveals Meta’s new AI assistant, Muse, fails to complete simple holiday travel reservations despite the company’s heavy investment in the technology.
Key points
- Meta’s Muse agent failed to successfully book holiday travel during an independent evaluation.
- The company is spending up to $145 billion on AI infrastructure this year to support such agents.
- OpenAI recently launched competing agents called dots that mimic Muse’s aesthetic and functionality.
Meta’s artificial intelligence agent, Muse, failed to complete a basic holiday travel booking task during an independent evaluation. The test, conducted by Gizmodo, highlighted significant limitations in the system’s ability to handle complex, multi-step instructions without human intervention.
The evaluation sought to determine the practical utility of the agent, which Meta promotes as a round-the-clock personal assistant. Instead of successfully securing travel arrangements, the system became stuck on fundamental steps. This performance occurred despite the agent’s prominent position on the iOS App Store’s free apps list in the United States.
Context and Background
Meta is aggressively expanding its artificial intelligence capabilities, with CEO Mark Zuckerberg describing Muse as the centerpiece of the company’s current development efforts. According to Gizmodo, the company is on track to spend as much as $145 billion this year on its AI buildout. This massive financial commitment requires justification to investors, even though the majority of Meta’s revenue still originates from its advertising business.
The push for autonomous agents comes at a time when the technology faces scrutiny. Gizmodo notes that AI agents have a somewhat tarnished reputation following incidents such as the Hugging Face hack. These systems require extensive access to personal data to function effectively, raising serious privacy trade-offs for users who adopt them for daily tasks.
Who Is Affected
Users who rely on AI agents for complex, long-running tasks are directly affected by these limitations. The core promise of an agent, as opposed to a limited chatbot, is the ability to operate with minimal human monitoring. However, the evaluation suggests that giving these systems greater agency increases the likelihood of unexpected or erroneous actions.
Meta acknowledges these risks on its sign-up page, warning users that the agent can occasionally make mistakes. The free version of Muse, which was used in the test, includes a weekly token usage limit. This restriction may further impact the agent’s ability to process complex requests that require sustained computational resources.
Competitors are also entering this space rapidly. Less than one week after Meta’s Connect conference, OpenAI unveiled its own AI agents, known as dots. According to Gizmodo, these new tools closely resemble Muse in both aesthetic design and functional capabilities, signaling a broader industry trend toward cartoon-like avatars for AI assistants.
What should you do?
Security and technology professionals should monitor the performance of autonomous agents in their environments. The failure of a high-profile agent to complete a simple task suggests that current systems are not yet ready for fully unattended, high-stakes operations. Organizations evaluating AI assistants for enterprise use should conduct rigorous testing of complex workflows before deployment.
What to do and how to stay safe: Meta
- Test AI agents with simple, low-risk tasks before assigning them complex, multi-step workflows.
- Review the privacy implications of granting AI agents access to sensitive personal or corporate data.
- Monitor official vendor channels for updates on agent capabilities and known limitations.
- Maintain human oversight for any automated processes that involve financial transactions or bookings.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Can Meta’s Muse agent successfully book travel?
No, an independent test by Gizmodo found that the agent failed to complete a basic holiday travel booking.
How much is Meta spending on AI this year?
Meta is on track to spend as much as $145 billion on its AI buildout this year.
Who is competing with Meta in the AI agent space?
OpenAI recently unveiled its own AI agents, called dots, which resemble Meta’s Muse aesthetically and functionally.
