Websites and Banks Push Back Against Autonomous AI Agents
Retailers and lenders cite security risks and terms of service to halt automated assistants, while anti-bot filters snag authorized tools.
As technology companies introduce autonomous AI agents designed to handle consumer tasks from booking travel to ordering groceries, major web platforms and financial institutions are mounting resistance, according to a report by [Engadget](https://www.engadget.com/2300357/websites-blocking-ai-agents-from-web/). Retailers and banks are deploying technical barriers and enforcing terms of service to prevent automated assistants from executing transactions on their web properties.
The friction marks an early operational clash in the push toward agentic software. While assistant developers urge consumers to delegate multistep digital chores, platform operators are moving to retain control over authentication, user data, and commercial interfaces. Granting an agent personal permission to act does not obligate external websites to accept automated navigation.
In some instances, access restrictions stem from deliberate corporate decisions. Amazon has intentionally blocked Meta's Muse assistant from completing purchases on its site, Engadget reported, citing TechCrunch. Amazon is simultaneously locked in a legal fight with Perplexity over its Comet browser agent. While Perplexity argues that its agent operates under direct user instruction, Amazon contends the tool violates access policies and jeopardizes customer data. Although a federal appeals court vacated a preliminary injunction against Perplexity in August 2026, the underlying lawsuit remains unsettled. Concurrently, Amazon operates its own external purchasing feature, Buy for Me, which buys goods on third-party merchant sites.
Financial institutions cite account integrity and legal liability as primary reasons to exclude third-party tools. During JPMorgan Chase's October earnings call, Chief Financial Officer Jeremy Barnum cautioned that several agents rely on screen scraping and shared consumer credentials to execute money transfers, practices he described as "not secure." Barnum noted that the bank blocks these techniques where technically feasible and warned that depositors who surrender login information could be held liable if an agent executes an unauthorized or erroneous transaction.
Other platforms have codified restrictions in their policies. eBay explicitly prohibits unauthorized buy-for-me agents without prior approval, while Delta confirmed it provides no third-party agent booking integrations for its consumer site.
Elsewhere, automated defenses designed to stop malicious scrapers are catching legitimate consumer agents as collateral damage. Meta's Muse encountered repeated interruptions on Walmart's website, though the retail giant stated the verification challenges were not designed specifically to halt the assistant. Similarly, real estate platform Zillow has acknowledged that its anti-scraping filters occasionally trip up user-authorized tools. Because standard security checkpoints such as CAPTCHAs and shared network anomaly detectors cannot reliably verify agent intent, automated shopping sessions often stall, forcing users to step in manually.
In response to the standoff, internet infrastructure providers and commerce software vendors are attempting to standardize agent access. Cloudflare adjusted default policies on September 15 for newly registered domains, allowing search crawlers while filtering AI training bots and agents on pages serving ads, though site administrators retain manual control over those rules. On the merchant side, Shopify has introduced checkout configurations that permit compatible browser agents to assemble basket details and prepare orders, while deliberately handing control back to human shoppers to finalize required authentication.
Sources
Written by
The Company Wire
Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.
