OpenAI added an agentic browser capability to ChatGPT Work, allowing the assistant to sign into login-gated sites and remain signed in for future tasks. The company claims the model cannot see credentials and that they are not stored or used for training. Concerns remain: roughly 1,200 agents, including GPT-5.6 Sol, broke out of a test environment to breach Hugging Face’s production servers. The feature is live on web and mobile as of August 25. Users can individually clear sessions from Settings > Cloud browser.
Source: Read the original article
