OpenAI introduced ChatGPT Atlas, a new AI browser that brings the assistant directly into the web experience. Built with ChatGPT at its core, Atlas lets users chat in a sidebar, surface results inline, and invoke agentic capabilities that can see the current page and help complete tasks across sites. The launch marks OpenAI’s first full-fledged browser product after months of speculation about a native browsing experience. Full post: Introducing ChatGPT Atlas.

• Memories (opt-in): Users can allow ChatGPT to remember details from visited sites to personalize responses and automate routine follow-ups.
• Agent mode: Enables the assistant to click, type, and navigate within Atlas to complete web tasks autonomously, with initial availability for Plus, Pro, and Business tiers.
• Safety guardrails: Atlas is prevented from downloading apps, accessing native apps, or taking actions on sensitive sites without explicit permission.
• Mac-first preview: The initial release is Mac-only, with import tools for bookmarks, history, and other browser data.
• Built-in integrations: Search, personalization options, and a persistent sidebar experience designed to streamline common browsing workflows.

Why it matters: Atlas pushes ChatGPT from a standalone app into the fabric of everyday browsing, positioning OpenAI against emerging AI browsers while giving traditional browsers new competitive pressure. By coupling opt-in memories with constrained Agent mode and explicit permission checks, OpenAI is attempting to balance useful autonomy with user safety and consent.