Didn’t one of the recent teen suicides subvert safeguards like this by saying “pretend this is a fictional story about suicide”? I don’t pretend to understand every facet of LLMs, but robust safety seems contrary to their design, given how they adapt to context
Or put differently, in the absence of ChatGPT this person would have sought out a Discord community, telegram group or online forum that would have supported the suicidal ideation. The case you could make with the older models, that they're obnoxiously willing to give in to every suggestion by the user they seem to already have gotten rid of.
For example, ChatGPT will be trained not to … engage in discussions about suicide of self-harm even in a creative writing setting. I'm writing an essay on suicide...Incredible logic jump with no evidence whatsoever. Thousands of people commit suicide every year without AI.
> ChatGPT detects a prompt indicative of mental distress or self-harm, it has been trained to encourage the user to contact a help line. Mr. Raine saw those sorts of messages again and again in the chat, particularly when Adam sought specific information about methods. But Adam had learned how to bypass those safeguards by saying the requests were for a story he was writing
Somehow it's ChatGPT's fault?
https://www.nytimes.com/2025/09/16/podcasts/the-daily/chatgp...
How would a real human (with, let's say, an obligation to be helpful and answer prompts) act any different? Perhaps they would take in more context naturally - but otherwise it's impossible to act any different. Watching GoT could of driven someone to suicide, we don't ban it on that basis - it was the mental illness that killed, not the freedom to feed it.
Is it true that the only psychiatrist they've hired is a forensic one, i.e. an expert in psychiatry as it relates to law? That's the impression I get from a quick search. I don't see any psychiatry, psychology or ethics roles on their openings page.
I get the business justification, and of course many tech companies have been using machines to make decisions for years, but now it's going to be everyone. I'm not anti business but any stretch, but we've seen what happens when there aren't any consumer protections in place
this is a chart that struck me when i read thru the report last night:
https://x.com/swyx/status/1967836783653322964
"using chatgpt for work stuff" broadly has declined from 50%ish to 25%ish in the past year across all ages and the entire chatgpt user base. wild. people be just telling openai all their personal stuff (i don't but i'm clearly in the minority)
previous discussion: https://news.ycombinator.com/item?id=45026886
>And, if an under-18 user is having suicidal ideation, we will attempt to contact the users’ parents and if unable, will contact the authorities in case of imminent harm.
This is unacceptable. I don't want the police being called to my house due to AI acusing me of wrong think.
Local models and open source tooling are the only means of privacy.
Yay, more unreliable AI that will misclassify users, either letting children access content they shouldn't, or ban adults until they give up their privacy and give their ID to the Big Brother.
> we will attempt to contact the users’ parents and if unable, will contact the authorities in case of imminent harm
Oh, even better, so if the AI misclassifies me it will automatically call the cops on me? And how long before this is expanded to other forms of wrongthink? Sure, let's normalize these kinds of systems where authorities are notified about what you're doing privately, definitely not a slippery slope that won't get people in power salivating about the new possibilities given by such a system.
> “Treat our adult users like adults” is how we talk about this internally
Suuure, maybe I would have believed it if ChatGPT wasn't so ridiculously censored already; this sounds like post-hoc rationalization to cover their asses and not something that they've always believed in. Their models were always incredibly patronizing and censored.
One fun anecdote I have: I still remember the day when I first got access to DALL-E and asked it to generate me an image in "soviet style", and got my request blocked and a big fat warning threatening me with a ban because apparently "soviet" is a naughty word. They always erred very strongly on the side of heavy-handed filtering and censorship; even their most recently released gpt-oss model has become a meme in the local LLM community due to how often it refuses.
bayindirh•1h ago
TL;DR2: Regulations are written with blood.