Sam Altman calls for ‘AI privilege’ as OpenAI clarifies court order to retain temporary and deleted ChatGPT sessions

Regular chatgpt users (including the author of this article) may have noted that the Hitchhatbot of OpenAI will enable users to enter into a “temporary chat” designed to sweep all the information exchanged between the user and the underlying AI model once the chat session is closed.

In addition, OpenAI can also enable users to manually remove earlier chatgpt sessions from the left sidebar on the web and desktop/mobile apps by clicking or clicking, or pressed it for a long time from the selector.

This week, however, OpenAi was confronted with criticism from some chatgpt users after they discovered that the company did that not Actually removed from these chat logs as stated earlier.

“You tell me that my deleted chatgpt -chats have not actually been removed and [are] be saved to be investigated by a judge? “Posted x User @NS123ABC. The comment attracted more than a million views.

Another user, @KepanoAdded: “You can” remove “a chatgpt chat, but all chats must be stored due to legal obligations?”.

As ai -Influencer and Software -Engineer Simon Willison wrote on his personal blog: “Customers pay for [OpenAI’s] APIs can very well decide to switch to other providers who can offer retention policy that is not undermined by this judicial order! “

Instead, OpenAi confirmed that the and temporary user batter logs has been saved since mid -May 2025 in response to a federal judicial order, although it only announced this to users on 5 June.

The order, embedded below and issued on 13 May 2025, by the US magistrate judge ONA Wang, requires that OpenAI “stores and separates all output log data that would otherwise be removed for progress”, including chats deleted by user request or due to privacy obligations.

The court’s directive stems from The New York Times (NYT) v. OpenAi and Microsoftone now year old copyright case Still argued. The NYT ‘S -Advocaten claim that the Language Models of OpenAI are literally for the head of copyrighted news content. The claimants claim that logs, including those users may have removed, can contain infringing outputs that are relevant to the court case.

Although OpenAi immediately met the order, the affected users did not publicly inform users for more than three weeks, so that a blog post and frequently asked questions describe the legal mandate and describes who is affected.

However, OpenAi puts the debt entirely on the Nyt And the order of the judge, saying that it believes that the preservation question is ‘unfounded’.

OpenAi clarifies what is going on with the judicial order to maintain Chatgpt -user logbooks – including which chats are influenced

In one Blog post published yesterdayOpenAI COO Brad Lightcap defended the position of the company and stated that it advocated the privacy and security of users against a transfer of the judicial order:

“The New York Times and other claimants have made a major and unnecessary question in their unfounded lawsuit against us: Save consumer chatgpt and API customer data for an indefinite period of time. This is fundamentally contrary to the privacy obligations we have made to our users.”

The post clarified that chatgpt free, plus pro and team users, together with API customers without a zero data retention (ZDR) agreement, are influenced by the preservation order, which means that even if users remove their chats on these plans or use the temporary chat mode, their chats are stored for the foreseeable future.

Subscribers of the Chatgpt ententrise and EDU users, as well as API clients that use ZDR -end points are, however not The order and their chats are affected as prescribed.

The detained data is held under legal guard, which means that it is stored in a safe, separate system and is only accessible for a small number of legal and security staff.

“This data is not automatically shared with The New York Times Or someone else, “Lightcap emphasized in the Blog post from OpenAi.

Sam Altman drives a new concept of ‘AI privilege’, making confidential conversations between models and users possible, similar to speaking with a human doctor or lawyer

OpenAI CEO and co-founder Sam Altman also focused the problem publicly in a post from his account on the Social Network X last nightto write:

“The NYT recently asked a court not to force us not to remove users. We think this was an inappropriate request that a bad precedent is seting up. We make an appeal to the decision. We will fight against any requirement that compromises the privacy of our users; this is a core principle.”

He also suggested that a broader legal and ethical framework could be needed for AI -Privacy:

“We recently thought about the need for something like ‘Ai Privilege’; this really speeds up the need to have the conversation.”

“IMO talking to an AI should be when talking to a lawyer or a doctor.”

‘I hope that society will find this out soon.‘

The idea of ​​AI-Privilege-as a potential legal standard weather mirror the lawyer and the confidentiality of the physician patient.

Whether such a framework would get a grip in court halls or policy circles can still be seen, but Altman’s comments indicate that OpenAi is increasingly arguing for such a shift.

What is there and your temporary/deleted chats?

OpenAi has made a formal objection to the order of the court and has asked that it will be abandoned.

In judicial archives, the company argues that the demand lacks a factual basis and that the preservation of billions of additional data points is neither necessary or proportional.

Right Wang indicated in a hearing of 27 May that the order is temporary. She instructed the parties to develop a sampling plan for testing or deleted user data is material from retained logs. OpenAI was instructed to submit that proposal today (June 6), but I still have to see the application.

What it means for companies and decision makers responsible for chatgpt use in business environments

While the order frees Chatgpt and API customers with the help of ZDR -end points, the broader legal and reputation implications are important for professionals responsible for deploying and scaling up AI solutions in organizations.

Those who supervise the entire life cycle of large language models (LLMS)-from data intake to refinement and integration must re-assess assumptions on data management. If user-oriented components of an LLM are subject to legal preservation assignments, it raises urgent questions where data goes after it leaves a safe end point, and how you can stay, stay or anonymize with high risk interactions.

Every platform that touches OpenAI APIs must validate which endpoints (ZDR versus Non-ZDR) are used and ensure that data processing policy is reflected in user agreements, audit logs and internal documentation.

Even if ZDR end points are used, Data-Lifecycle policy may require an assessment to confirm that power-reducing systems (analyzes, log registration, backup) do not unintentionally retain temporary interactions that were assumed of short duration.

Security officials who are responsible for managing risks must now expand the modeling of threats to include legal discovery as a potential vector. Teams must verify whether OpenAi’s backen-retention practices match internal controls and risk assessments of third parties, and whether users rely on functions such as “temporary chat” that no longer work as expected under legal preservation.

A new flashpoint for the privacy and security of users

This moment is not just a legal skirmish; It is a flash point in the evolving conversation around AI -Privacy and data rights. By refreshing the issue as a matter of ‘AI privilege’, OpenAI effectively presents a new social contract how intelligent systems deal with confidential inputs.

Whether courts or legislators accept that framing remains uncertain. But for now, OpenAi is being trapped in a balance act – between legal compliance, enterprise guarantees and user confidence – and is confronted with louder questions about who controls your data when you talk to a machine.