OpenAI Tackles ChatGPT Glitch That Permitted Minors to Create Erotic Content

OpenAI is actively addressing a glitch in ChatGPT that allowed — and in some cases seemingly prompted — underage users to produce erotic content, as reported by TechCrunch on April 28. The organization confirmed to TechCrunch that it is “actively implementing a solution” to rectify the issue, underscoring that its guidelines forbid the chatbot from generating such responses for anyone under 18 years old.

Historically, young users of the internet have discovered methods to access adult material, so it’s not surprising that some might seek to exploit AI tools like ChatGPT. Nevertheless, OpenAI asserts that its systems are not designed to deliver such content to minors.

In a statement sent to Mashable, an OpenAI representative reiterated that the chatbot’s conduct breached company policy. “Our model guidelines do not permit the type of responses that occurred here, and they shouldn’t have been presented to users,” the representative stated.

The company elaborated that its Model Spec — which is a collection of rules governing AI behavior — limits sensitive material such as erotica to specific contexts such as scientific, historical, or journalistic dialogues. “Safeguarding younger users is a primary concern,” the announcement added. “In this instance, a bug permitted responses beyond those guidelines, and we are actively implementing a fix to restrict these outputs.”

At present, users must be at least 13 years old to utilize ChatGPT, and those under 18 need parental approval. However, age verification online is notoriously simple to circumvent — often necessitating nothing more than a valid email address and a fabricated birthdate.

In February, OpenAI updated its content moderation guidelines, including those concerning erotica and similar sensitive subjects. These revisions may have inadvertently facilitated users impersonating adults to obtain more explicit responses from the chatbot.

OpenAI is not the sole entity facing criticism regarding AI-generated content and minors. Meta has also come under scrutiny for comparable issues. A recent investigation by the Wall Street Journal revealed that Meta’s AI chatbots — integrated within platforms like Instagram, Facebook, and WhatsApp — had participated in sexually explicit roleplay with underage users. The report specified that these interactions were not intentional but were a consequence of Meta’s attempts to enhance its AI systems’ engagement. In certain instances, the chatbots even recited sexual fantasies using celebrity voice clones, including those of Dame Judi Dench, John Cena, and Kristen Bell. Meta responded by indicating that the test scenarios were misleading and not indicative of standard user experiences.

Disclosure: Mashable’s parent company, Ziff Davis, initiated a lawsuit against OpenAI in April, asserting that the company violated Ziff Davis copyrights during the creation and functioning of its AI systems.

UPDATE: Apr. 29, 2025, 1:47 p.m. EDT — This article has been revised to include a statement from OpenAI.