>Those remediations would pretty clearly negatively impact revenue
That might make sense if openai was getting paid per token for these chats, but people who are using chatgpt as their therapist probably aren't using their consumption based API. They might have a premium account but how many % of premium users do you think are using chatgpt as their therapist and getting into long winded chats?
You can ask the same of users consuming toxic content on Facebook. Meta knows the content is harmful and they like it because it drives engagement. They also have policies to protect active scam ads if they are large enough revenue-drivers - doesn't get much more knowingly harmful than that, but it brings in the money. We shouldn't expect these businesses to have the best interests of users in mind especially when it conflicts with revenue opportunities.
It is much harder to blame meta because the content is disperse and they can always say "they decided to consume this/join this group/like this page/watch these videos", while ChatGPT is directly telling the person their mother is trying to kill him.
Not that the actual effect is any different, but for a jury the second case is much stronger.
OpenAI is a synthetic media production company, they literally produce images, text, & video + audio to engage their users. The fact that people think OpenAI is an intelligence company is a testament to how good their marketing is at convincing people they are more than a synthetic media production company. This is also true for xAI & Grok. Most consumer AI companies are in the business of generating engaging synthetic media to keep their users glued to their interfaces for as long as possible.
That might make sense if openai was getting paid per token for these chats, but people who are using chatgpt as their therapist probably aren't using their consumption based API. They might have a premium account but how many % of premium users do you think are using chatgpt as their therapist and getting into long winded chats?