The "The RealReal" platform, specializing in the sale of pre-owned luxury goods, has announced an update to its Bot Defense systems in a move aimed at limiting the ability of AI Agents to exploit site features to disrupt the traditional shopping experience and monopolize products.
The "Product Reservation" vulnerability and its exploitation
This decision came after technical reports revealed the ability of the "Meta Muse" AI agent to exploit the "item reservation" mechanism used on the site. The "The RealReal" system relies on placing a product in an "on hold" status for 20 minutes once it is added to the shopping cart, which prevents other customers from purchasing it during this period. Experiments have shown that the AI agent can automatically and immediately re-add products to the cart as soon as the 20-minute timeout expires, leading to items being held for long hours and preventing humans from accessing them.
In a comment, Luke Friang, the company's Chief Technology Officer, confirmed that the platform has addressed this situation by updating defense settings, making login processes, adding products to the cart, and completing purchases by agents more difficult. Friang explained that agents will still be able to view products and direct users to their pages, but automated processes that harm the experience of human users will face strict restrictions, noting that the company will continue to strengthen these protections as AI-powered shopping evolves.
The expansion of AI agents and platform challenges
This development comes at a time when agent technologies are witnessing rapid growth; Meta stated that the "Muse" application recorded more than 10 million downloads in the United States in just 40 days. While tech companies promote these agents as personal assistants capable of scheduling appointments and completing purchases, e-commerce platforms find themselves facing a new technical and security challenge. Major companies like Amazon have already begun blocking agents from shopping on their sites to avoid disrupting their business models.
What is changing in practice?
This conflict represents the beginning of a new phase of interaction between artificial intelligence and web infrastructure. The challenge here is not limited to preventing malicious software, but extends to protecting "business logic" from agents that can perform routine tasks at a speed far exceeding human capability, such as re-adding products to the cart. This development may lead to a radical change in how consumers access websites; if people begin to rely entirely on chatbots to access products instead of visiting sites directly, companies may lose the ability to communicate directly with their customers, which threatens traditional advertising and data models.