OpenAI’s ChatGPT Work can now sign into your accounts and operate while you sleep

OpenAI launched a cloud browser feature for ChatGPT Work on August 25, 2026, enabling the AI to log into password-protected websites autonomously and continue tasks even when the user's device is off. Available to ChatGPT Plus, Pro, and Business/Enterprise subscribers, the system uses session cookies and never exposes credentials to the model, but high-impact actions still require user confirmation. The feature removes the biggest barrier to automation, though it raises security concerns about centralized session tokens.

Photo: Tima Miroshnichenko / Pexels OpenAI’s ChatGPT Work can now sign into your accounts and operate while you sleep The new cloud browser feature lets the AI authenticate into password-protected sites autonomously, raising questions about where convenience ends and risk begins. OpenAI just gave ChatGPT the ability to log into websites on your behalf, enter credentials through a secure form, and keep working even after you close your laptop. The feature, which launched on August 25, 2026, represents the company’s most aggressive push yet toward fully autonomous AI agents. Available to ChatGPT Plus, Pro, and Business/Enterprise subscribers, the update equips ChatGPT Work with a dedicated cloud browser that can authenticate into password-protected services and maintain those sessions across multiple tasks. How the login system actually works Users enter their credentials through a secure form that OpenAI says the underlying model never sees. The AI doesn’t store your passwords either. Instead, it relies on session cookies, the same tokens your browser uses to remember that you’re logged in, which persist until they expire or you manually clear them. This means ChatGPT can start a task on one site, pivot to another application, pull data from a third, and return to the first without asking you to re-enter your password each time. The AI operates from OpenAI’s cloud infrastructure, completely independent of your local device. Your laptop could be powered off in a drawer somewhere while ChatGPT is filing expense reports across three different platforms. OpenAI is careful to note that high-impact actions still require explicit user confirmation. The company hasn’t disclosed exactly what qualifies as “high-impact.” The update also introduces event-based triggers for Scheduled Tasks, meaning ChatGPT Work can now respond to specific conditions rather than just running on a timer. If a particular email arrives, a spreadsheet gets updated, or a dashboard metric crosses a threshold, the AI can kick off a predefined workflow automatically. The productivity promise and the security question ChatGPT Work was initially introduced earlier in 2026 as a system built for executing complex, multi-step workflows rather than just answering questions. The login capability removes what had been the single biggest obstacle to true automation: the fact that most useful work happens behind a login screen. Before this update, any AI-powered workflow that needed to interact with authenticated services would hit a wall. The user had to either stay present to punch in credentials or rely on API integrations, which are limited to services that offer them. But the security implications are equally obvious. Session cookies, while standard web technology, are also a well-known attack vector. If OpenAI’s cloud infrastructure were compromised, those persistent sessions could theoretically grant attackers access to every service a user has authenticated through the system. OpenAI asserts that credentials are not retained or used for training purposes, but the session tokens themselves represent a concentrated target. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .