Main AI developers deploy autonomous agents

The main developers of artificial intelligence (AI), including OpenAI, Meta, Google, and xAI, have deployed autonomous web agents to perform online tasks such as reservations and purchases. However, commercial platforms in the retail, travel, and banking sectors are actively working to exclude these tools.

Meta launches personal AI assistant Muse

At the beginning of September 2026, Meta launched a personal AI assistant called Muse, intended for people aged 18 and older seeking help with their daily tasks, such as managing their schedules, making purchases, and transforming "their long-term goals into action plans." Muse operates on a dedicated and secure virtual machine that hosts both the agent and the user's data.

OpenAI introduces Dots for background tasks

Then, at the end of September, OpenAI launched "Dots," personal AI agents that continue to perform your tasks in the background even after the chat window is closed. Announced during the DevDay in San Francisco, they operate on GPT-6 Astra and made their appearance only a few weeks after Meta's "Muse" topped the app download rankings.

Commercial platforms block AI agents

According to a report from The Verge, commercial platforms in the retail, travel, and banking sectors are actively working to exclude these tools. Systems such as Meta's Muse, OpenAI's Dots, xAI's Grok, and Instinct are now encountering technical obstacles, access barriers related to licenses, and explicit account prohibitions.

Websites restrict access to AI agents

Websites are closing the door on AI agents. According to The Verge, Yelp only allows agents that pay for its data licensing. Delta blocks agents from booking tickets. JPMorgan and Charles Schwab block agents from accounts, for security reasons. Amazon keeps fighting agent shopping, from Perplexity to Meta's Muse. And many agents simply trip the normal anti-bot systems built for scrapers and scammers.

Retailers split on AI agent access

Shopify opened its storefronts to agents, and Meta partnered directly with retailers like Walmart and Best Buy. So the web is splitting: sites that publish a door for agents, and sites that don't. This divide reflects a growing tension between platforms that see value in integrating AI tools and those that prioritize security and advertising revenue.

Guidelines for building AI agents

Four rules worth keeping if you build agents: 1. If your agent works by pretending to be a human in a browser, you are building on someone else's patience. One anti-bot update and the feature is gone. 2. Build on the doors that exist. Official APIs, MCP servers, partner programs. Less coverage today, much more reliable.

Legal and technical challenges persist

One place where open chains have it easier: a smart contract endpoint has no captcha. The rules are in the code, the same for a human and for an agent. For everything else, the agents that last will be the ones that got invited. In many sectors, the main online services have implemented explicit restrictions to prevent autonomous agents from performing tasks on behalf of users.

Anti-bot systems limit agent effectiveness

The growing wave of restrictions in the travel, financial, and online retail sectors reflects a general rejection within the industry of uncoordinated web-based automated agents. The technical obstacles faced by autonomous software stem both from agent-specific policies and existing anti-bot infrastructures deployed on the web.

Platforms and developers clash over automation

Even when platforms maintain official partnerships with agents, automated tools frequently trigger standard anti-abuse heuristics. Users running Meta's Muse have reported account blocks by Google for suspicious activity, while buyers have encountered payment failures on Walmart despite an official partnership between Meta and the distribution chain.

Economic conflict shapes the future of AI agents

The blocking of AI agents occurs in a delicate context. The tensions around autonomous agents reveal an economic conflict between the utility of automation and the advertising-based business models of platforms. Amazon has stated in its legal documents that third-party purchasing agents must respect sellers' consent, while Perplexity argued that the retailer opposed automated purchases because these agents eliminate human exposure to sponsored ads.

Uncertain future for AI agents

Developers have promised that software personal assistants would handle tedious digital tasks and integrate into daily digital routines, but the widespread blocking limits their capabilities in the real world. Chris Canal, CEO of EquiStamp, stated that if platforms actively prevent automated navigation, the perceived value of major developers like Anthropic and OpenAI diminishes, as agents can accomplish fewer tasks.

Security measures burden ordinary users

Regular internet users and businesses bear the immediate operational costs of this growing digital impasse. As web platforms erect defensive barriers to keep automated assistants at bay, human visitors face increasingly intrusive verification systems. Chris Canal warned that security measures would make standard navigation more burdensome for the general public.

Retail partnerships offer limited relief

The future of autonomous AI agents on the web remains uncertain. The commercial sector has split into two opposing camps regarding access to autonomous agents. While Amazon, Delta Air Lines, and Yelp erect defensive barriers, other retailers have opted for commercial alliances. Meta announced that refusing an automated personal agent equated to rejecting the paying customer hidden behind it and thus established partnerships with Walmart, Gap, Sephora, and Best Buy to allow Muse to carry out tasks effectively.