Artificial Intelligence

OpenAI Launches 'ChatGPT Agent' for Automated Web Tasks

Published JUL 18 2025

Story

Above: OpenAI's ChatGPT logo. Image credit: Vincent Feuray/Hans Lucas/AFP via Getty Images

The Facts

OpenAI launched "ChatGPT Agent" on Thursday, reportedly enabling the AI assistant to complete complex, multi-step tasks using its virtual computer, including browsing websites, running code, and creating editable documents such as spreadsheets and presentations.
Openai
Tom's Guide
The new agent combines capabilities from OpenAI's previous Operator web browsing tool and Deep Research feature, allowing users to request tasks such as calendar management, competitor analysis, meal planning, and financial modeling through natural language commands.
Openai
TechCrunch
ChatGPT Agent is available to Pro, Plus, and Team subscribers, with Pro users receiving 400 monthly tasks and other paid tiers getting 40 tasks per month. Enterprise and Education users will gain access in the coming weeks.
Openai
MacRumors
The system requires explicit user permission before taking consequential actions, such as making purchases or accessing sensitive websites. It includes built-in safeguards, including "Watch Mode," which pauses execution when users navigate away from sensitive sites.
Openai
Verge
OpenAI classified the ChatGPT Agent as having "High Biological and Chemical capabilities" under its Preparedness Framework, implementing enhanced safety measures, including real-time content monitoring and refusal training, to prevent potential misuse.
Openai
Tom's Guide
Openai
The agent scored 41.6% on Humanity's Last Exam and 27.4% on FrontierMath, outperforming previous models and competing systems, such as Microsoft Copilot in Excel.
Openai
Tom's Guide
Safe

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

This breakthrough represents the future of productivity, transforming ChatGPT from a simple chatbot into a capable digital assistant that actually gets work done. The agent can handle tedious tasks like research, scheduling, and document creation while users focus on higher-value activities. With robust safety measures and user control, this technology delivers on the long-promised vision of AI that truly assists rather than just converses.

Maginative

Tech

Techno-skeptic narrative

The launch raises serious concerns about privacy and security risks that OpenAI admits it cannot fully predict. CEO Sam Altman himself warned families to treat this as "cutting edge and experimental" technology unsuitable for high-stakes uses. The potential for malicious actors to exploit these agents through prompt injection attacks, combined with the unprecedented access to personal data, creates dangers that may outweigh the convenience benefits.