Aside from the possible introduction of artificial general intelligence (AGI), AI agents, autonomous processes that you can instruct to perform complex tasks for you on your computer, will be perhaps the biggest new AI feature in 2025. Agents could be essential for turning your mobile phone into a true AI assistant, capable of doing whatever you ask it without you needing to get involved.
OpenAI has been teasing us with the release of its first AI agent, called Operator, for a while now, but the latest code leak suggests that it could arrive very soon and on the Mac.
A new leak on X from Tibor Blaho claims to have revealed evidence that OpenAI’s Operator agent is coming to the ChatGPT Mac app. Tobor has discovered hidden options to define shortcuts for the desktop launcher to “Toggle Operator” and “Force Quit Operator,” which might indicate that you might need a quick way to shut it down if it gets out of control!
Confirmed – the ChatGPT macOS desktop app has hidden options to define shortcuts for the desktop launcher to “Toggle Operator” and “Force Quit Operator” https://t.co/rSFobi4iPN pic.twitter.com/j19YSlexASJanuary 19, 2025
Tobor also claims to have found code in the browser version of ChatGPT that references Operator with references to an “Operator System Card Table,” “Operator Research Eval Table,” and “Operator Refusal Rate Table.” The last entry indicates that perhaps the Operator fails to perform the tasks it is asked to do enough to require a refusal rate.
OpenAI website already has references to Operator/OpenAI CUA (Computer Use Agent) – “Operator System Card Table”, “Operator Research Eval Table” and “Operator Refusal Rate Table”Including comparison to Claude 3.5 Sonnet Computer use, Google Mariner, etc.(preview of tables… pic.twitter.com/OOBgC3ddkUJanuary 20, 2025
Automating daily tasks
Recently, one of the founders of OpenAI, Wojciech Zaremba, slammed rival Anthropic in a post on X for releasing its AI agent without the necessary safety precautions in place. His post read:
“Anthropic — just released a computer-using agent without any safety mitigations. I can only imagine the negative reactions if OpenAI made a similar release”.
Where does the gap between perception and reality on AGI company safety come from?https://t.co/hu2UBxwkSW — Elon is very vocal about safety, but so far, no one at https://t.co/hu2UBxwkSW works on safety.Anthropic — just released a computer-using agent without any safety…December 14, 2024
It’s an AI agent’s ability to integrate into your daily computer tasks that makes it such a big step forward for AI and has the potential to change how we interact with our devices entirely. Just imagine if you didn’t have to book hotel rooms yourself, pay bills, or even write code.
Obviously, a lot of work is going to need to be done before people will trust an AI agent to perform such tasks autonomously, and privacy will be a key issue.
