Major Websites Block AI Agents from Automated Shopping and Online Transactions
Artificial intelligence agents such as Meta's ‘Muse’ and Perplexity's ‘Comet’ are rapidly gaining popularity for automatically completing complex tasks on behalf of users, including online shopping and flight bookings. However, these AI agents are facing significant obstacles when attempting to access websites directly to collect information or purchase products according to users' instructions.
Primarily citing commercial competition, data scraping, and security concerns, multinational giants such as Amazon, Walmart, and JPMorgan are increasingly restricting access for these virtual agents.
Security and Financial Transaction Complications
Financial institutions view restricting AI agents from platforms involving direct financial transactions as a fundamental security requirement. Recently, Jeremy Barnum, Chief Financial Officer (CFO) of leading US bank JPMorgan Chase, said that some AI agents obtain users' confidential login credentials to perform automated tasks, creating serious cybersecurity risks. The bank has warned that users will bear full responsibility for any unintended financial transactions carried out through such unauthorised agents.
Meanwhile, global e-commerce giant Amazon has accused AI agent technologies such as Perplexity's of violating its access policies and putting customers' personal data at risk.
However, Amazon itself offers a similar service called ‘Buy for Me’, which can access other websites and complete purchases automatically. Analysts believe this double standard in the technology and e-commerce industries clearly exposes the underlying battle over commercial interests.
Confusion Between CAPTCHA Systems and Ordinary Bots
However, AI agents are not always blocked intentionally. In many cases, legitimate AI agents are also prevented from completing tasks because of conventional security systems designed to stop malicious bots and automated data collectors.
Standard ‘CAPTCHA’ or human-verification systems on popular platforms such as Walmart and Zillow are still unable to accurately distinguish between real human users and AI agents. As a result, ordinary users often find their AI assistants unable to complete tasks at the final stage, forcing them to return to the screen and finish the work themselves.
To address this problem, cybersecurity companies such as Cloudflare are working to introduce new protocols. These would allow website owners to distinguish between ordinary AI bots engaged in data theft and legitimate AI agents acting on behalf of human users, enabling them to grant appropriate access permissions.
//DBTech/BMT/OR//
