OpenAI Launches ChatGPT Agent with Operator and Deep Research Integration

Prime Highlights

OpenAI introduces the ChatGPT Agent, a powerful software capable of executing multi-step activities like scheduling, bookings, and web research independently.
The Agent integrates Operator and Deep Research, along with integrated security functionality for task authorization and monitoring.

Key Fact

Launched today for Pro, Plus, and Team customers, with broader availability planned for Enterprise and Education later this summer.
Framed with strict safety measures for one-way commands to offer control and clarity to users.

Key Background

OpenAI has made an announcement of releasing the ChatGPT Agent, a landmark development in AI task automation. This new instrument merges two advanced technologies: Operator, which performs human-like interactions such as browsing, clicking, and file handling, and Deep Research, which conducts and compiles thorough insight collection independently within minutes. By combining these technologies, the ChatGPT Agent presents an integrated solution for managing both digital research and effective task execution with ease.

The Agent’s capabilities extend far beyond prompting, as it can perform activities such as scheduling appointments, shopping, creating spreadsheets, generating presentations, and even sorting email and scheduling calendars on its own. Demonstrations have revealed the Agent reserving tables, buying things online, and constructing formal research reports. Unlike other AI systems, it can leave ongoing work behind, refine itself based on user input in real-time, and verify actions prior to committing irreversible acts, such as sending an e-mail or making a payment.

OpenAI has been eager to highlight control and safety. The ChatGPT Agent operates at high levels of capability, with hard controls to prevent it from being abused. Features like “Watch Mode” enable user monitoring while dealing with sensitive tasks, and robust protections filter out malicious or manipulative content on the internet. CEO Sam Altman emphasized that the tool is “experimental and bleeding-edge,” reminding users to avoid high-risk operations or sensitive information processing in the absence of close oversight.

The product is initially released to Pro, Plus, and Team users, with usage limitations such as 400 agent messages per month for Pro members. Enterprise and Education will follow later this year, although release in certain jurisdictions, like the EU, remains pending. With task times ranging from 15 to 30 minutes, the Agent promises faster execution of intricate workflows than manual attempts.