Lockdown Mode started with enterprise customers. Now it’s reportedly yours.
OpenAI has reportedly completed the Lockdown Mode rollout to personal ChatGPT accounts, including Free, Plus, Pro, and self-serve Business plans, as of June 8, 2026. The hub first covered Lockdown Mode’s architecture and broader rollout announcement on June 6; this update confirms the personal account tier completion, per Infosecurity Magazine’s reporting. The primary source URL for the June 8 completion date isn’t confirmed in independently fetched content, use the “reportedly” framing literally.
What Lockdown Mode does
The mechanism is deterministic, not probabilistic. Security analyst Simon Willison’s June 6 analysis explains the design clearly: Lockdown Mode blocks outbound network requests at the system level. That’s different from filtering what the model says, it’s cutting the network paths that would let data leave ChatGPT in the first place. Prompt injection attacks that instruct ChatGPT to send data to an external URL can’t succeed when there’s no outbound network access to exploit. The defense is structural, not model-level.
What you lose when you enable it
The trade-off is real. Per OpenAI’s Help Center documentation of the feature, enabling Lockdown Mode disables live connectors, write actions, and third-party tool integrations. If your ChatGPT workflow includes web browsing, Zapier integrations, or any plugin that reads from or writes to external systems, those stop working in Lockdown Mode. That’s the feature, not a bug, the point is to sever the paths data can travel out of the session.
What to Watch
For individual users with straightforward text-generation workflows, the trade-off is minimal. For users running integrated workflows through ChatGPT’s tool ecosystem, enabling Lockdown Mode effectively turns ChatGPT into a standalone reasoning engine, powerful, but isolated.
The question OpenAI hasn’t answered
Lockdown Mode is opt-in. OpenAI hasn’t disclosed whether it will ever become default-on for any account tier. That matters because the security benefit only applies to users who know the feature exists and actively enable it. Most personal ChatGPT users won’t read this brief. They won’t see the toggle. The exfiltration defense is real for the security-aware minority that turns it on, for everyone else, the attack surface is unchanged.
Don’t expect OpenAI to make this default-on soon. Default behavior changes in consumer products require broader UX validation than security features warrant alone. Enterprise plans moved to GA in February; personal plans reportedly completed June 8. The arc of the rollout is toward broader availability, not toward defaulting on.
What to watch
The next signal is whether OpenAI publishes updated documentation explicitly confirming the June 8 personal GA date, the hub’s prior June 6 verification remains the strongest basis for the feature’s existence. Watch for Help Center updates at openai.com. If you’re a Plus or Pro user, the toggle is in your ChatGPT settings under security options, check for it now.
For the detailed breakdown of what Lockdown Mode blocks and how it fits into agentic AI security architecture, the hub’s June 6 deep-dive covers the architecture in full.