United Kingdom, Nov. 12 -- Former OpenAI product safety lead Steven Adler has voiced serious concerns over the company's decision to reintroduce erotic content generation for verified adults, warning that the move may be premature given ChatGPT's ongoing safety challenges.

In a new interview with WIRED, Adler recounted a 2021 internal crisis at OpenAI, when his team discovered that users were prompting GPT-3 to generate erotic and sometimes disturbing content.

He said: "An uncomfortable amount of this traffic was devolving into all sorts of sexual fantasies.

"Even if you weren't intending to go to an erotic role-play place or certain types of fantasies, the AI might steer you there."

The issue, he explained, revealed a deeper problem:...