Doe v. OpenAI Raises Pseudonymity and Safety Questions

Doe v. OpenAI alleges that an ex-boyfriend, driven into a delusional spiral by ChatGPT, generated dozens of fake psychological reports about the plaintiff and disseminated them to her network, then escalated to death threats. The plaintiff filed under a pseudonym; her lawyers say the pseudonym will likely be temporary because the defendant is "dangerous" and "paranoid," and public naming could increase the risk of violence. The filing has prompted skepticism about whether pseudonymity is necessary or effective.

The complaint claims harassment materially facilitated by outputs from ChatGPT. For practitioners this raises practical evidentiary and discovery questions: model outputs, prompt histories, account metadata, hosting logs, and any private communications generated or distributed are potentially discoverable from both the user and the provider. Courts will weigh whether plaintiff identity redaction is needed while preserving the ability to subpoena logs and other digital evidence from OpenAI or intermediaries.

Legal bases courts consider for pseudonymity:

• •Demonstrable threat of physical violence or stalking to the plaintiff or third parties
• •Severe privacy interests or stigma that would cause irreparable harm if names are public
• •Specific, contemporaneous facts showing how disclosure would increase risk

Courts balance these against the presumption of open judicial records and the public interest in transparent adjudication. Temporary pseudonymity is more likely when the factual record shows imminent danger; it is less likely where secrecy would impede defendant identification or meaningful discovery.

This case sits at the intersection of generative AI harms and civil procedure. It tests how traditional doctrines about pseudonymity, open courts, and discovery adapt when alleged wrongdoing uses ChatGPT to scale and manufacture deceptive content. The decision could shape tactical choices for plaintiffs and defendants in AI-related torts: whether to seek sealed filings, how to frame safety claims, and how courts order production of model interaction logs.

Whether the court grants temporary pseudonymity and the scope of any protective orders, how it structures discovery of ChatGPT-related records, and whether the court articulates a new framework for balancing safety and transparency in AI-enabled harassment cases.