|
|
|
|
August 21, 2025
|
Hackers Infiltrate Alleged North Korean Operative’s Computer, Leak Evidence of...
|
|
August 21, 2025
|
Ecosia Proposes Unusual Stewardship Model for Google Chrome
|
|
August 21, 2025
|
OpenAI Presses Meta for Evidence on Musk’s $97 Billion Takeover Bid
|
|
August 15, 2025
|
ChatGPT Mobile App Surpasses $2 Billion in Consumer Spending, Dominating Rivals
|
|
|
OpenAI Launches ChatGPT Agent: A Powerful New Tool to Automate Digital Tasks
July 17, 2025
OpenAI has unveiled ChatGPT Agent, a general-purpose AI assistant designed to go far beyond answering questions — it can now take action on your behalf. This new tool enables users to delegate complex computer-based tasks using simple natural language commands, making ChatGPT feel more like a digital coworker than just a chatbot.
A Smarter, More Capable AI Assistant
Available to Pro, Plus, and Team subscribers, ChatGPT Agent rolls out as a powerful upgrade within the existing ChatGPT interface. By activating “agent mode”, users unlock a range of advanced capabilities, including:
Automatically navigating calendars
Generating editable slides and presentations
Running code
Planning meals and making purchases
Analyzing competitors and creating reports
OpenAI has integrated features from earlier internal tools — like Operator, which mimics clicking and navigating websites, and Deep Research, which compiles structured reports from multiple sources — into one unified experience.
How It Works
ChatGPT Agent can access third-party apps and services through ChatGPT connectors, allowing it to interact with platforms like Gmail and GitHub. It also has access to a command-line terminal and can use APIs to fetch or manipulate data, making it ideal for both technical users and productivity-focused professionals.
One example OpenAI gives: simply ask the agent to "plan and buy ingredients to make Japanese breakfast for four," and it will search, decide, and execute the steps necessary to complete the task — all without further user input.
Record-Breaking Benchmark Scores
Backed by a cutting-edge model, ChatGPT Agent delivers some of the best performance yet across industry benchmarks:
41.6% on Humanity’s Last Exam (pass@1) — more than double the previous top scores from earlier models
27.4% on FrontierMath with tool access — compared to 6.3% from the last best-performing model
These results suggest the new agent is not only smarter but also more reliable at completing diverse and complex tasks.
Built with Safety in Mind
With increased capabilities comes heightened responsibility. OpenAI has added a suite of real-time safety measures, including:
A biology classifier that screens prompts and outputs for content related to biological threats
A double-layer filter to detect and block potentially harmful outputs
Memory disabled for the agent to prevent misuse, particularly through techniques like prompt injection
These safeguards are part of OpenAI’s broader Preparedness Framework, which aims to mitigate risk while allowing useful features to evolve.
What This Means for the Future of AI Agents
ChatGPT Agent is OpenAI’s boldest move yet toward making autonomous AI agents mainstream. For years, AI tools have promised to offload tedious digital tasks — but in practice, they often fell short. Now, OpenAI claims to have built a system that is significantly more capable, safer, and ready for wider use.
Whether it’s helping a busy professional automate research, supporting a developer through hands-free coding assistance, or simply managing daily tasks like meal planning, ChatGPT Agent marks a big step forward in everyday AI utility.
The big question now: Will users trust an AI to act on their behalf — and will the agent consistently deliver?
|
|
|
Sign Up to
Our Newsletter!
Get the latest news in tech.
|
|
|
