OpenAl’s ChatGPT Agent casually clicks through “I am not a robot” verification test | “This step is necessary to prove I’m not a bot,” wrote the bot as it passed an anti-Al screening step.
https://arstechnica.com/information-technology/2025/07/openais-chatgpt-agent-casually-clicks-through-i-am-not-a-robot-verification-test/
4 Comments
Submission statement: “ChatGPT Agent is a feature that allows OpenAI’s AI assistant to control its own web browser, operating within a sandboxed environment with its own virtual operating system and browser that can access the real Internet. Users can watch the AI’s actions through a window in the ChatGPT interface, maintaining oversight while the agent completes tasks. The system requires user permission before taking actions with real-world consequences, such as making purchases.
Recently, Reddit users discovered the agent could do something particularly ironic.”
My first thought was, “Lmao.”
My second was, “Shit, captchas are going to become even more obnoxious.”
Captcha hasn’t been about stopping all bot access for a long time. It’s been able to be bypassed by bot scripts long before the whole ai boom. It’s about filtering out the much more widespread, broader bots that flood everything.
This isn’t ground breaking despite what news sites want you to think.
When I’ve run Agent, it gets snagged on Cloudflare. Even when I assume control of the browser, I can’t get past Cloudflare. I’m using Netscape Navigator 3.1, so maybe that’s why? (Kidding about my browser, not kidding about Cloudflare)