OpenAI Launches ChatGPT Atlas, Sparks Privacy Concerns

OpenAI this week released ChatGPT Atlas, a Chromium-based AI browser that embeds ChatGPT to provide on-page summarization and planning, persistent “browser memories,” an experimental agent mode that can act on webpages, and tools to import bookmarks and passwords from other browsers. OpenAI says Atlas will not store extremely sensitive items like passwords or medical records and provides memory controls. Reviewers and privacy advocates report those controls can be non‑intuitive and say clearing history does not always remove stored memories. Cybersecurity experts warn Atlas expands attack surfaces — from prompt‑injection to agents accessing logins or payment information — which could enable data leaks, fraud or account takeovers and will require continual security work. The launch positions OpenAI as a direct challenger in an escalating AI‑browser arms race and raises broader questions about content discovery, data access and how browsing data may be used to train models.