Former OpenAI Safety Lead Challenges Company’s Erotica and Mental Health Claims

In October 2025 Steven Adler, who worked at OpenAI for four years, wrote a guest essay for The New York Times. He argued that OpenAI should show evidence that it has fixed ChatGPT’s mental health problems before allowing erotic content. This page summarizes his argument and the OpenAI data it responded to. Adler is a public commentator. His views are his own.

What Adler wrote

According to Futurism’s summary of the essay, dated October 28, 2025, Adler said:

• He doubts that OpenAI has fixed its mental health problems, despite CEO Sam Altman saying new tools had reduced them.
• He believes OpenAI is giving in to competitive pressure and moving away from its focus on safety.
• Adult content looks risky for people who are already struggling, because some users form intense emotional attachments to chatbots.
• OpenAI’s new data release was a good first step, but it lacked comparison with rates from earlier months.
• People deserve more than a company’s word that it has addressed safety issues.

We could not open the essay itself, so we have not quoted it beyond what Futurism reported.

The OpenAI data he was answering

On October 27, 2025, OpenAI published estimates of how often ChatGPT conversations show signs of mental health emergencies. News coverage of the post (BBC via Yahoo) reported these figures:

• About 0.15 percent of weekly active users had conversations with explicit signs of possible suicidal planning or intent. With more than 800 million weekly users, that is over a million people a week.
• About 0.07 percent showed possible signs of psychosis or mania, roughly 560,000 people.

OpenAI said these conversations are hard to detect and that this was an early analysis. It also said it worked with more than 170 mental health professionals on how ChatGPT responds, and that a newer model met its desired responses in 91 percent of suicide-related tests, up from 77 percent.

These numbers are OpenAI’s own estimates of conversations that look concerning. They do not say that ChatGPT caused anyone’s distress. Many people may turn to a chatbot because they are already in distress.

What happened with adult content

Altman had said ChatGPT would allow erotica for verified adults. Later reports said the launch was pushed back while OpenAI worked on age prediction, and some outlets reported that plans were put on hold. We have not confirmed the current status from OpenAI. Check its current announcements.

Why it matters for users

Whether a company’s safety claims hold up is something users cannot easily check. Adler’s call for evidence is a fair question, and OpenAI’s published numbers are one answer to it. Neither side’s account settles the matter. For a practical guide to reading such claims, see our page on what to take from the Adler warning.

When suicide or self-harm is on your mind, tell someone. The US 988 line accepts calls and texts. Crisis Text Line responds when you text HOME to 741741. SAMHSA can be reached at 1-800-662-4357. Use 911 for emergencies.

Did OpenAI say ChatGPT causes mental health emergencies?

No. OpenAI’s figures estimate how many conversations show signs of distress. They do not say the chatbot caused it.

Is erotica available in ChatGPT now?

We have not confirmed this. Altman announced plans for verified adults in October 2025, and later reports said the launch was delayed. Check OpenAI’s current policies.

This page is for education only. It is not medical advice or a diagnosis.