OpenAI has introduced a new tool on its ChatGPT platform to offer more support to users, such as doing the heavy lifting for wedding planning or online shopping.
The feature, ChatGPT agent, is capable of "handling complex tasks from start to finish", San Francisco-based OpenAI said on Thursday. In other words, it can process long, detailed queries and return results in the same manner – and then some.
The new tool has accelerated the generative artificial intelligence race. It is the latest of OpenAI's tools as the company seeks to and assert its dominance in the crowded field, distinguishing itself from major companies such as Microsoft, Google, Anthropic and Elon Musk's Grok, as well as upstarts including DeepSeek and Qwen.
What can ChatGPT agent do?
In a detailed video, OpenAI chief executive Sam Altman and his team flaunted ChatGPT agent's capabilities, with Mr Altman pointing out that "people wanted a unified agent that could go off" on its own.
ChatGPT agent, according to OpenAI, can handle even the most complex, long and detailed tasks. As an example, Mr Altman's team asked ChatGPT to plan a wedding, complete with suggestions for five outfits that matched the theme, gifts under $500 and hotel options on Booking.com.
After seven seconds, ChatGPT agent replied that it was looking into it, in addition to more queries. In this case, it wanted to know the exact date of the wedding, which was not mentioned earlier.
It resulted in detailed suggestions and options in response to the request. While this is not surprising, the highlight is that it did not miss a beat and answered everything in the query.
How does it work?
ChatGPT agent uses its "own computer", a suite of tools that includes a visual browser that interacts with the internet through a graphical-user interface, a text-based browser for simpler reasoning-based web queries, a terminal and direct API access.
It is also capable of connecting to apps including Gmail and developer platform Github, so ChatGPT can find information that is relevant to queries and use them in responses, OpenAI said.
In addition, users can take over the visual browser by logging in on a website, giving them full control and having ChatGPT go deeper in its "research and task execution".
"Giving ChatGPT these different avenues for accessing and interacting with web information means it can choose the optimal path to most efficiently perform tasks," OpenAI added.
Is ChatGPT agent free?
Users can activate ChatGPT agent by clicking on the tools button under ChatGPT's prompt line and selecting "agent". However, you can only use it if you are subscribed to ChatGPT's paid tiers – Pro, Plus or Team, which cost $20, $200, and $25 or $30 a month, respectively.
This was to be expected, because advanced tools normally are not available for free.
The introduction of ChatGPT agent began on Thursday, with Pro users getting first the crack at it. The Plus and Team tiers will gain access to it "over the next few days", while Enterprise and Education users will have to wait. However, it still will not be available for the European Economic Area and Switzerland.
Not risk-free
This is the first time users can ask ChatGPT to take action on the internet and doing so introduces new risks, "particularly because ChatGPT agent can work directly with your data, whether it’s information accessed through connectors or websites that you have logged it into via takeover mode", OpenAI said.
"ChatGPT agent’s expanded tools and broader user reach mean its overall risk profile is higher," it added.
A key safeguard in ChatGPT agent is that it limits access to certain tasks, from personal actions such as sending emails, which require user oversight, to more sensitive, "high-risk tasks" including bank transfers, which ChatGPT agent would refuse to do.
In any case, such tasks potentially or do involve personal and confidential user data that, in theory, could be misused. "Overall, we expect continued improvements to ChatGPT agent’s efficiency, depth and versatility over time, including more seamless interactions as we continue to adjust the amount of oversight required from the user to make it more useful while ensuring it’s safe to use," OpenAI said.
