OpenAI announces deeper ChatGPT integration with Google Chrome tabs

What happened

OpenAI announced a deeper integration between ChatGPT and the Google Chrome web browser this week [1]. This new feature includes the ability to hand control over browser tabs to the AI [1].

While the integration has been demonstrated to work in normal ChatGPT conversations, a user reported technical difficulties when attempting to use these browser control features through a “dot” [2]. According to a report from a user on a Windows computer, the “Chrome Integration” worked successfully during a standard chat session, but the same workflow failed when using the dot to start local browser tasks [2].

Specific technical issues noted in the report include:

• An error code MXC 0x80070003 occurred when attempting to read browser tabs [2].
• After a server restart on October 14, a task startup attempt returned an HTTP 401 Unauthorized error [2].
• During this failed attempt, the computer was reported as connected, attached, and authorized for tasks, and ordinary chat messages could still be read, but no task ID was returned and no browser tool ran [2].
• The user noted that while Chrome control was successfully demonstrated, they have not confirmed if control works for the Brave browser [2].
• The user checked the Chrome integration and extension component files and paths, finding them to be present, but a fresh control session still failed with the same error [2].
• Restarting the desktop application did not resolve the error [2].

The user is seeking clarification on whether the difference in behavior between “Chrome Integration” in a normal chat and the cua_repl in a dot is expected, and has asked for diagnostic steps or a confirmed fix for the 401 error [2].

Who is affected

• Users of ChatGPT on Windows computers who use the Google Chrome browser [2].
• Users attempting to use browser control tasks through a “dot” [2].
• Users of the Brave browser, though successful control has not been confirmed [2].