OpenAI deactivates PRO account after 28-hour Codex session in VS Code

A viral Reddit post highlights a growing concern among heavy AI users: OpenAI's account deactivation policy may be targeting legitimate power users. The user, a solo developer, had their PRO account (recently renewed for $200/month) deactivated after running a Codex agent in VS Code for approximately 28 hours. The /goal task involved writing articles, generating images and assets, and building small tools for their own niche websites. They explicitly checked OpenAI's policy violation list and confirmed they did not engage in any prohibited activities such as spam, hate speech, or intellectual property abuse.

The appeal was rejected, leaving the user without access to their active context and a full day of work. In panic, they continued from a second Plus account to avoid losing progress. They speculate that the long-running session with large repo context, repeated retries, and compacting errors may have triggered an automated security flag. This incident raises questions about OpenAI's ability to distinguish between abusive automation and legitimate high-volume usage, especially for developers using Codex for extended autonomous workflows. The community response has been mixed, with some suggesting better communication from OpenAI and others warning about the risks of relying on a single account for long-running agents.