ChatGPT's new locked chats sit behind a 4-digit PIN, but are not end-to-end encrypted

What happened
OpenAI began rolling out "locked chats" for ChatGPT on Oct. 20, 2026. The feature moves selected conversations out of regular history, pinned chats and search into a separate section opened with a four-digit PIN, or Face ID or Touch ID on supported phones. OpenAI's own support page says the lock is not end-to-end encryption, and it does not change how long the company retains the chats or its model-training settings.
What we know
- Locked chats are hidden from chat history, pinned conversations and chat search, and are grouped in a dedicated "Locked Chats" section, per TechCrunch and Engadget.
- Access requires a four-digit PIN. On supported iPhone and Android devices, biometric authentication such as Face ID can be used instead, per Engadget and TechCrunch.
- OpenAI says locking does not create end-to-end encryption or an encrypted vault, change how long it retains a chat, or alter the "Improve the model for everyone" setting, per TechCrunch. Engadget reports the company states this on a newly published support page.
- Resetting the PIN permanently deletes all locked chats and uploaded files in the section, and OpenAI Support cannot recover them, per Engadget and TechCrunch.
- With Memory enabled, Locked Chats keeps its own separate memory that does not feed into regular Memory, per TechCrunch. Information ChatGPT remembered before a chat was locked is not removed.
- Locking is unavailable for temporary, archived and group chats, chats created from shared links, chats with scheduled tasks or meeting recordings, and Codex conversations, per TechCrunch and Engadget.
What we don't know yet
- Exact rollout timing per account is unclear. Both outlets say the release is gradual and may take days to reach some users.
- Plan coverage differs slightly between reports. Engadget lists Free, Go, Plus and Pro, while TechCrunch also lists Business, Enterprise and Edu and says the feature is not available on ChatGPT for Healthcare. The OpenAI support page was not among the sources reviewed.
- Neither source says how locked chats are stored on OpenAI's servers or whether they are treated differently in legal requests. OpenAI says only that the lock does not change retention or its legal obligations, per Engadget.
Why it matters
The feature guards against someone seeing a conversation on a shared or unattended screen, but it gives no protection against OpenAI's retention or training policies, and a forgotten PIN means permanent loss of the chats. Users and workplaces that treat "locked" as "private from the company" would be wrong under OpenAI's own documentation.
Claims
- ConfirmedOpenAI began rolling out locked chats for ChatGPT on Oct. 20, 2026.
- ConfirmedLocked chats are opened with a four-digit PIN, or with biometric authentication such as Face ID on supported phones.
- ConfirmedLocked chats are not end-to-end encrypted, according to OpenAI.
- ConfirmedLocking a chat does not change how long OpenAI retains it, according to OpenAI.
- ConfirmedResetting the PIN permanently deletes all locked chats and their uploaded files, and OpenAI Support cannot recover them.
- ConfirmedLocked Chats has its own separate memory that does not update regular ChatGPT Memory.
- ConfirmedCodex conversations cannot be locked.
- ConfirmedLocked chats are available on ChatGPT Free, Go, Plus and Pro plans.
- ConfirmedLocked chats are also available on ChatGPT Business, Enterprise and Edu plans but not on ChatGPT for Healthcare.
Related coverage
Why you can trust this story
75%Source map · 2 outlets / 2 articles
- Engadget
- Article 1engadget.comreport
- TechCrunch
- Article 1techcrunch.comreport
How this credibility score is calculated
- Source reliability
- Corroboration
- Primary source
- Atom-verified claims
- No contradiction
- Settled
- Claim attribution
- AI disclosure
Atom-verified claims matched cited source text. Weights are fixed and explainable.
Human accountability
- 2 independent origins / 2 sources
- Drafted by the Newsmesis judgment agent (agent-cli); human approval required before publishing.
Verification ledger · 9
- Claim: OpenAI began rolling out locked chats for ChatGPT on Oct. 20, 2026.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locked chats are opened with a four-digit PIN, or with biometric authentication such as Face ID on supported phones.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locked chats are not end-to-end encrypted, according to OpenAI.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locking a chat does not change how long OpenAI retains it, according to OpenAI.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Resetting the PIN permanently deletes all locked chats and their uploaded files, and OpenAI Support cannot recover them.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locked Chats has its own separate memory that does not update regular ChatGPT Memory.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Codex conversations cannot be locked.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locked chats are available on ChatGPT Free, Go, Plus and Pro plans.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
- Claim: Locked chats are also available on ChatGPT Business, Enterprise and Edu plans but not on ChatGPT for Healthcare.Status: ConfirmedVerification trailVerifiedChecked by verification engine
At least one extracted atom matched cited source full text.
Update log · 0
AI accelerates. High-risk topics wait for a human editor.