Amazon blocks Meta's AI agent, Muse, for making purchases without formal identification
Read more
Tecnoblog
tecnoblog.net

Amazon blocks Meta's AI agent, Muse, for making purchases without formal identification

Amazon has restricted access to Muse, a new artificial intelligence agent developed by Meta, on its e-commerce platform. According to GeekWire, this restriction was implemented because the retailer observed that the assistant was making purchases and accessing account information without providing formal identification.

For Amazon, such conduct violates the platform's terms of service and may pose risks to consumer security. Muse, launched this month, was designed to perform tasks autonomously, such as filling out forms and completing orders on behalf of the user.

Initially, Amazon sought an agreement, asking Facebook's controller to voluntarily remove the site from the robot's reach, but discussions did not progress. Without reaching a consensus, users who tried to use the Meta assistant for purchases last weekend were confronted with an error message, warning that continuous access by an unauthorized agent violates the terms of use signed by customers.

Amazon maintains that Muse operates by accessing order histories and processing transactions without obtaining explicit consent from the store. A company spokesperson emphasized that third-party applications with purchasing capabilities must operate transparently to ensure a safer and more reliable experience.

In response, Meta guaranteed that the tool is completely secure and was developed prioritizing privacy. The tool's creator stated that the agent does not have visibility into consumers' passwords or payment methods. According to Meta, credentials entered by the user in the browser are stored in an isolated virtual machine, allowing the AI to use them without direct exposure.

The owner of WhatsApp also stressed that any sensitive action, such as finalizing a financial transaction, requires mandatory prior user approval and verification by a second internal monitoring agent called Sentinel.

It is worth noting that both companies maintain an established partnership since 2023 for integrated sales on Instagram and signed a high-value contract in April for processing AI loads on Amazon's own cloud infrastructure.

This is not the first case where the retail giant imposes limits on AI traffic. Amazon had already prevented other autonomous purchasing systems from Google and OpenAI, in addition to being involved in litigation against the startup Perplexity due to the use of the AI browser Comet to track store pages.

In parallel, Amazon is launching its own AI solutions, such as the recent Alexa for Shopping. However, the company highlights that its own assistants identify themselves when accessing third-party websites, giving brands the option to reject automated traffic on their domains.

Similar stories

Meta launches Muse, an autonomous AI agent capable of managing emails, purchases, and travel bookings
Read more
olhardigital.com.br

Meta launches Muse, an autonomous AI agent capable of managing emails, purchases, and travel bookings

Meta has introduced Muse, an artificial intelligence (AI) agent designed to operate as a personal digital assistant. This tool has the capability to autonomously interact with various applications and websites on behalf of the user, executing tasks such as sending emails, booking trips, and making online purchases.

This launch signals one of the first initiatives by a major technology corporation to make an AI agent available to the general public. Unlike conventional chatbots, these systems are designed to complete actions on their own, rather than being limited to responding to direct commands.

Muse can be accessed through conversations, functioning like a chatbot, where it receives instructions to perform various activities. The tool is accessible via a dedicated application as well as WhatsApp, a platform owned by Meta itself.

The agent can be linked to the company's existing services, including Facebook and Instagram. With user permission, it can also access third-party services and applications such as Spotify, Ticketmaster, Shopify, Gmail, and OpenTable. This integration allows Muse to use data from multiple accounts to gain a deeper understanding of the user and execute tasks on their behalf.

A distinguishing factor of Muse compared to other AI agents is its deep integration with the Meta ecosystem. The agent can consult information from Facebook and Instagram, using this data to offer a more personalized experience to its users. However, this functionality raises privacy concerns. Meta assures that all personal data collected by Muse, as well as the agent itself, is stored securely on its servers and cloud computing infrastructure.

It is important to note that Muse is intended exclusively for adults. Users have the option to customize the tool with their own name and avatar.

More details about Muse

Muse is considered one of the first significant consumer-facing products developed by Meta Superintelligence Labs, a division created by CEO Mark Zuckerberg to accelerate the company's efforts in the AI competition. Zuckerberg has already classified Muse as Meta's next big breakthrough, explaining that the agent was built to work '24 hours a day, seven days a week on their behalf,' helping the user achieve goals and improve aspects such as personal life, health, relationships, and finances.

Meta has invested billions of dollars in developing foundational AI models and building data centers to compete with other companies, such as Google, OpenAI, and Anthropic. So far, the company's strategy has shown mixed results. Meta's AI smart glasses have achieved millions of sales but have also raised privacy concerns.

In July, the company temporarily suspended its AI image generator on Instagram due to criticism related to copyright and privacy. The Muse agent is powered by Muse Spark, an AI model launched by Meta in April, which is the first model developed by the division under the leadership of Alexandr Wang, the company's AI director. However, Muse Spark still performs inferiorly to the leading models from Anthropic and OpenAI. Meta, however, is working on a more robust system, internally codenamed Watermelon, which may be launched next month.

The company plans to expand Muse's functionalities in the coming months, including integration with Meta's smart glasses, which feature a camera and voice assistant. Although the Muse app is free, there are usage limits. To increase these limits, users can opt to pay $20 (approximately R$ 104) or $100 (approximately R$ 520) monthly.

With the introduction of Muse, Meta is inserting itself more directly into an AI category that promises to take over some of the tasks currently performed by users themselves. The distinction lies precisely in the level of autonomy: instead of just conversing, the agent can access services, perform specific actions, and complete tasks on behalf of the user.

Popular