OpenAI could be on the verge of releasing a groundbreaking AI tool that could potentially control your computer and carry out tasks on your behalf. This development is linked to the much-anticipated Operator tool, which has been rumoured to be in the works for some time.
Tibor Blaho, a software engineer known for accurately leaking AI product details, claims to have discovered fresh evidence of Operator’s progress. Earlier reports from outlets like Bloomberg hinted that Operator would be an “agentic” system, capable of automating tasks such as writing code or managing travel bookings.
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/j19YSlexAS
— Tibor Blaho (@btibor91) January 19, 2025
Evidence points to a January release
According to a report by The Information, OpenAI is targeting a January release for Operator. Over the weekend, Blaho uncovered a hidden code in OpenAI’s ChatGPT macOS client, suggesting that the tool is nearing completion. The code includes options for shortcuts like “Toggle Operator” and “Force Quit Operator,” which remain disabled for now.
Blaho also reported that OpenAI’s website contains unpublished references to Operator, including performance tables that compare it to other computer-using AI systems. While these tables may be placeholders, they offer a glimpse into Operator’s potential capabilities and limitations.
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"
— Tibor Blaho (@btibor91) January 20, 2025
Including comparison to Claude 3.5 Sonnet Computer use, Google Mariner, etc.
(preview of tables… pic.twitter.com/OOBgC3ddkU
The leaked benchmarks indicate mixed results. On OSWorld, a platform that evaluates AI performance in simulated computer environments, OpenAI’s “Computer Use Agent” (CUA)—believed to be the foundation for Operator—achieved a score of 38.1%. This surpasses the performance of a similar model by Anthropic but falls well below the human benchmark of 72.4%.
In web-based benchmarks, Operator performed exceptionally on WebVoyager, which measures an AI’s ability to navigate websites, but struggled on WebArena, another web interaction test. Specific tasks proved challenging; for example, Operator only managed a 60% success rate when setting up a virtual machine and a mere 10% success rate when creating a Bitcoin wallet.
Competitors and safety concerns
OpenAI’s move into the AI agent market comes amid growing competition from major players like Anthropic and Google. While still in their infancy, these AI-powered agents are being promoted as the next big advancement in artificial intelligence. Analysts at Markets and Markets predict that the AI agent market could grow to $47.1 billion by 2030.
However, the rise of AI agents has raised concerns about their reliability and safety. Leaked documents suggest Operators have undergone rigorous safety evaluations, including tests designed to prevent them from engaging in illicit activities or accessing sensitive personal data. Despite these efforts, some experts remain wary about the rapid development of such technology.
OpenAI has faced criticism for prioritising product releases over safety, with former employees and researchers expressing concerns about the potential risks. OpenAI co-founder Wojciech Zaremba recently criticised a competitor, Anthropic, for releasing an agent he claims lacks adequate safety features. “I can only imagine the negative reactions if OpenAI made a similar release,” Zaremba said in a post on X (formerly Twitter).
Looking ahead
OpenAI has yet to comment publicly on the leaks, and questions remain about how reliable and safe Operator will be in real-world applications. If the tool launches as planned, it could reshape how we interact with AI in everyday tasks, marking a significant step forward in developing autonomous systems.
As the tech world awaits further announcements, one thing is clear: the race to dominate the AI agent space is heating up, and Operator could be a game-changer.
