Introducing ChatGPT agent

Source: OpenAI Date: 2025-07-17 URL: https://openai.com/index/introducing-chatgpt-agent

Summary

OpenAI’s launch announcement for ChatGPT Agent, published July 2025 alongside the system card. This is the product-facing announcement — the system card covers safety; this covers the user experience and use cases. ChatGPT Agent is the consumer-facing autonomous mode that can browse the web, execute code, manage files, and interact with web applications across a multi-step task within a single session. Positioned as the agentic evolution of ChatGPT rather than a separate product.

Implications

The product-safety announcement pattern. Publishing the launch post and system card simultaneously is OpenAI’s mature launch playbook for high-risk products — capability announcement paired immediately with safety documentation, leaving less space for “they launched without thinking about harms” criticism. Compare to DALL-E 2’s staged rollout or the original ChatGPT launch which had no accompanying safety documentation.

Consumer agent as the market moment. ChatGPT Agent landing in July 2025 — concurrent with the GPT-5.x series and before the Workspace Agents enterprise product (April 2026) — establishes consumer autonomy before enterprise. This sequencing is interesting: typically enterprise deployments precede consumer for high-risk capabilities. OpenAI’s consumer-first approach reflects confidence in their safety work and competitive urgency.

Thread: agentic product evolution. Read alongside ChatGPT Agent System Card (July 2025), Operator (January 2025), Workspace Agents (April 2026), and the Responses API computer environment (March 2026) as the agentic product buildout sequence.

Watch: Adoption rate of agent mode vs. standard ChatGPT among Plus subscribers — the metric that determines how much OpenAI invests in the agentic surface going forward.