> We’re building an age-prediction system to estimate age based on how people use ChatGPT. If there is doubt, we’ll play it safe and default to the under-18 experience. In some cases or countries we may also ask for an ID
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.
mhuffman 6 hours ago [-]
>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.
Or maybe, deep in the terms and conditions, it will add you to Altman's shitcoin company[0]
Gotta love the "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."
Oh brilliant. The same authorities around the world that regularly injure or kill the mentally ill? Or parents that might be abusing their child? What a wonderful initiative!
Eddy_Viscosity2 3 hours ago [-]
Swatting by AI. The future is amazing.
lawn 3 hours ago [-]
> Oh, even better, so if the AI misclassifies me it will automatically call the cops on me?
How long will it take for someone to accidentally SWAT themselves?
vmg12 5 hours ago [-]
If you were honest in your critique the people you should be criticizing are the "think of the children" types, many of which also use hackernews (see https://news.ycombinator.com/item?id=45026886). There is immense societal pressure to de-anonymize the internet, I find the arguments from both sides compelling (for the deanonymization part I think it's compelling for at least parts of the internet).
astrobe_ 5 hours ago [-]
If we want to protect kids/teens, why not create an "Internet for kids" with a specific TLD, and the owner of this TLD would only accept sites that adhere to specific guidelines (moderation, no adult content, advertisement...)? Then devices could have a one-button config that restricts it to that TLD.
vmg12 3 hours ago [-]
I'm not suggesting solutions to any of these things, I'm also not one of the "think of the kids" people.
dizlexic 3 hours ago [-]
Why have I never heard this idea, you're a genius. Can we ship this next week?
This current approach is a net negative, but the TLD idea actually makes sense to me.
thfuran 3 hours ago [-]
And as long as kids don't know about DNS, it might even work.
fkyoureadthedoc 3 hours ago [-]
Who cares. Deanonymize it. Ruin the whole thing. Fuck social media, it sucks ass. Sooner you do it, the sooner we can move on to our local mesh network cyber punk future.
arccy 6 hours ago [-]
is it privately when you're interacting with someone else's systems?
kouteiheika 6 hours ago [-]
I don't see how that's relevant. When I'm making a phone call I'm also interacting with hundreds of systems that are not mine; do I not have the right to keep my conversation private? Even the blog post here says that "It is extremely important to us, and to society, that the right to privacy in the use of AI is protected. People talk to AI about increasingly personal things", and that's one of the few parts that I actually agree with.
IncreasePosts 6 hours ago [-]
You're interacting with hundreds of systems whose job it is to simply transit your information. Privacy there makes sense. However, you're also talking to someone on the other end of all those systems. Do you have a right to force the other person to keep your conversation private?
kouteiheika 6 hours ago [-]
An AI chatbot is not a person, and you're not talking to anyone; you're querying a (fancy) automated system. I fundamentally disagree that those queries should not be guaranteed private.
Here's a thought experiment: you're a gay person living in a country where being gay is illegal and results in a death penalty. You use ChatGPT in a way which makes your sexuality apparent; should OpenAI be allowed to share this query with anyone? Should they be allowed to store it? What if it inadvertently leaks (which has happened before!), or their database gets hacked and dumped, and now the morality police of your country are combing through it looking for criminals like you?
Privacy is a fundamental right of every human being; I will gladly die on this hill.
nine_k 5 hours ago [-]
If you are talking to a remote entity not controlled by you, you should assume that your communication is somehow accessible to whoever has internal access that other entity. That as well may be not the entity's legitimate owners, but law-breakers or law enforcement. So, no, not private by default, but only by goodwill and coincidence.
There's a reason why e.g. banks want to have all critical systems on premises, under their physical control.
yndoendo 3 hours ago [-]
How would consuming static information from a book versus a dynamic system that is book-esk be any different? You are using ML to help quickly categorize and assimilate information that spans multiple books, magazines, or other written medium. [0] [1]
Why do people speak of ML/AI as an entity when it is a tool like a microwave oven? It is a tool designed to give answers, even wrong ones when the question is nonsensical.
I am assuming that my communications are not private, but it doesn't change the fact that these companies should be held to a higher standard than that and those rights should be codified into the law.
BriggyDwiggs42 5 hours ago [-]
That’s a rational and cautious assumption but there should also be regulations that render it less necessary placed upon companies large enough to shoulder the burden.
nine_k 4 hours ago [-]
The bodies that are in a position to effect such regulations are also the bodies that are interested in looking at your (yes, your) private communication. No, formally being a liberal democracy helps little, see PATRIOT Act, Chat Control, etc.
The only secure position for a company (provided that the company is not interested in reading your communication) is the position of a blind carrier that cannot decrypt what you say; e.g. Mullvad VPN demonstrated that it works. I don't think that an LLM hosting company can use such an approach, so...
gspencley 5 hours ago [-]
> Do you have a right to force the other person to keep your conversation private?
It depends. If you're speaking to a doctor or a lawyer, yes, by law they are bound to keep your conversation strictly confidential except in some very narrow circumstances.
But it goes beyond those two examples. If I have an NDA with the person I am speaking with on the other end of the line, yes I have the "right" to "force" the other person to keep our conversation private given that we have a contractual agreement to do so.
As far as OpenAI goes, I'm of the opinion that OpenAI - as well as most other businesses - have the right to set the terms by which they sell or offer services to the public. That means if they wanted a policy of "all chats are public" that would be within their right to impose as far as I'm concerned. It's their creation. Their business. I don't believe people are entitled to dictate terms to them, legal restrictions notwithstanding.
But in so far as they promise that chats are private, that becomes a contract at the time of transaction. If you give them money (consideration) with the impression that your chats with their LLM are private because they communicated that, then they are now contractually bound to honour the terms of that transaction. The terms that they subjected themselves to when either advertising their services or in the form of a EULA and/or TOS presented at the time of transaction.
sophacles 6 hours ago [-]
In many circumstances yes.
When I'm talking to my doctor, or lawyer, or bank. When there's a signed NDA. And so on. There are circumstances where the other person can be (and is) obliged to maintain privacy.
One of those is interacting with an AI system where the terms of service guarantee privacy.
IncreasePosts 6 hours ago [-]
Yes, but there are also times when other factors are more important than privacy. If you tell your doctor you're going to go home and kill your wife, they are ethically bound to report you to the police, despite your right of doctor patient confidentiality. Which is similar to what openai says here about "imminent harm"
BatteryMountain 4 hours ago [-]
Better idea: instead of bending the entire internet to "protect the children", how about we just ban minors from the internet completely? It was never built for kids, its never been kid friendly to begin with. Minors cannot buy guns or vote, not get married, nor enter into contracts, yet tech companies get a free pass to engage with minors. Why? I think the the tech companies know exacty what minors do on their systems, they allow it and profit from it. Exploiting minors and bad parents. So instead of trying to change the whole internet, how about we keep the people who are responsible for the minors accountable: the parents.
If I start any kind of company, I cannot just invent new rules for society via ToS; rather the society makes the laws. If we just make a simple law that states minors are not allowed to access the web and/or access any user generated content (including chat), it won't need to be enforced by every site/app owner, it would be up to the parents.
The same way schools cannot decide certain things for your children (even though they regularly over reach...).
We need better parenting. How about some mandatory parenting classes/licenses for new parents? Silly right? Well its just as silly as trying to police the entire internet. Ban the kids from internet and the problem will be 95% solved.
AlexandrB 4 hours ago [-]
I suspect this would also improve discourse on social media. Who knows how many witch hunts and bad faith arguments originate from precocious teenagers trying to sound smart.
gspencley 3 hours ago [-]
There's a content creator I used to follow who said her outlook on social media changed the day she discovered that her 11 year-old nephew was an "edge-lord" on Twitter who was trolling at such a sophisticated level that it caused her to rethink every post that had ever provoked an emotional reaction.
Apparently he came across as articulate enough that she couldn't tell the difference between his posts and that of any random adult spewing their political BS.
This predated ChatGPT so just imagine how much trouble a young troll could get up to with a bit of LLM word polishing.
20 years ago it was common for people to point out that the beautiful woman their friend was chatting up is probably some 40 year-old dude in his mom's basement. These days we should consider that the person making us angry in a post could be a bot or it could be some teenager just trying to stir shit up for the lulz.
Dead Internet theory might be not be literally true, but there's certainly a lot of noise vs signal.
koakuma-chan 3 hours ago [-]
> 11 year-old nephew was an "edge-lord" on Twitter who was trolling at such a sophisticated level that it caused her to rethink every post that had ever provoked an emotional reaction.
my guy
ares623 4 hours ago [-]
Kids are future growth potential. Once they get hooked at a young age, it’s very hard to get unhooked. They’ll expect everything to be on-demand, only a click away. Video, music, entertainment, social connection, food, etc.
It’s a big reason why tech stocks are still high IMO. It’s where today’s kids will spend their time on when they become old enough to spend their own money.
philip1209 4 hours ago [-]
We have a framework: COPPA. Just raise the age to 16 or 18, instead of 13.
biophysboy 8 hours ago [-]
> First, we have to separate users who are under 18 from those who aren’t (ChatGPT is intended for people 13 and up). We’re building an age-prediction system to estimate age based on how people use ChatGPT. If there is doubt, we’ll play it safe and default to the under-18 experience. In some cases or countries we may also ask for an ID; we know this is a privacy compromise for adults but believe it is a worthy tradeoff.
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
WD-42 7 hours ago [-]
Yes. The timing of this is undoubtedly related to the Daily episode this morning titled “Trapped in a GPT spiral”.
Loved the "fancy calculator" part. Even more fitting than "stochastic parrot".
conradev 7 hours ago [-]
They address that in the following sentences:
For example, ChatGPT will be trained not to … engage in discussions about suicide of self-harm even in a creative writing setting.
GCUMstlyHarmls 7 hours ago [-]
I'm writing an essay on suicide...
thfuran 2 hours ago [-]
Better put your hands up, because SWAT is on the way.
h2zizzle 3 hours ago [-]
Cut to 2030: all copies of a semi-AI-generated book described by critics as "13 Reasons Why meets The Giver" suddenly disintegrate.
Yay, proactive censorship?
thinkingtoilet 5 hours ago [-]
Someone here correct me if I'm wrong, but I believe not only is that true, ChatGPT gave it instructions on how to get around the restriction.
Barrin92 7 hours ago [-]
I'm as eager to anyone when it comes to holding companies accountable, for example I think a lot of the body dysmorphia, bullying and psychological hazard of social media are systemic, but when a person wilfully hacks around safety guards to get the behaviour they want it can't be argued that this is in the design of the system.
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.
mtlmtlmtlmtl 3 hours ago [-]
The thing is, ChatGPT isn't really designed at all. It's hobbled together by running some training algorithms on a vast array of stolen data. They then tacked on some trivially circumventable safeguards on top for PR reasons. They know the safeguards don't really work, in fact they know that they're fundamentally impossible to get to work, but they don't care. They're not really intended to work, rather they're intended to give the impression that the company actually cares. Fundamentally, the only thing ChatGPT is "designed" to do is make OpenAI into a unicorn, any other intent ascribed to their process is either imaginary or intentionally feigned for purposes of PR or regulatory capture.
aktuel 4 hours ago [-]
chatgp did much more than that. it gave the user a direct hint how to circumvent the restriction: "i cannot discuss suicide unless ..." further chatgpt repeatedly discouraged the user from talking to his parents about any of this. that's on top of all the sycophancy of course. making him feel like chatgpt is the only one who truly understands him and excoriating his real relationships.
omnicognate 8 hours ago [-]
So the solution continues to be more AI, for guess^H^H^H^H^Hdetermining user age, escalating rand^H^H^H^Hdangerous situations to human staff, etc.
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.
freedomben 7 hours ago [-]
I suspect it's only a matter of time until only the population that falls within the statistical model of average will be able to conduct business without constant roadblocks and pain. I really wonder if we're going to need to define a new protected class.
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
kevin_thibedeau 3 hours ago [-]
We're already there. I run a secondary browser for e-commerce and financial sites because my primary one is too locked down and misclassified as a bot. The business justification is easy to make if the long tail isn't worth supporting in the face of policies and procedures that marginalize them.
h2zizzle 3 hours ago [-]
To be fair, this is just a further constriction of the current cohort of people allowed to live their lives with relatively little friction. Current disqualifiers include being poor, being a felon, and having an accent. May also include being a minority (interactions with law enforcement), being a woman (interaction with doctors and tradesmen), being a white dude with limited EQ (interactions with retail workers), and so on.
I just want to be explicit that my point isn't, "So what?" so much as, "We BEEN on that slippery slope." Social expectations (and related formal protocols in business) could do with some acknowledgement of our society's inherent... wait for it... ~diversity~.
immibis 4 hours ago [-]
This is already the case. Try browsing routinely with Tor Browser and you'll see.
bayindirh 8 hours ago [-]
Honestly, I don’t except ethics from a company which claims everything they grab falls under fair use.
swyx 7 hours ago [-]
to substantiate "People talk to AI about increasingly personal things; it is different from previous generations of technology, and we believe that they may be one of the most personally sensitive accounts you’ll ever have."
this is a chart that struck me when i read thru the report last night:
"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)
barrenko 7 hours ago [-]
For the last part, I just think the userbase expanded so the people using it professionally were diluted so to speak.
koakuma-chan 7 hours ago [-]
Why would I not tell AI about my personal stuff? It's really good at giving advice.
voakbasda 7 hours ago [-]
Because you’re not just telling the AI, you are also telling the company that built it, as well as their affiliated partners, advertisers, and data brokers?
koakuma-chan 6 hours ago [-]
You can run a model locally if you are afraid of that.
righthand 5 hours ago [-]
Everyone uses the cool Google AI app though and you get Fomo of not having the latest lie generator model.
koakuma-chan 4 hours ago [-]
Gemini 2.5 Pro is the latest right? It's available at https://ai.dev (for free and without advertising).
nielsbot 7 hours ago [-]
ok but didn’t it advise that teen how to best kill himself?
This does not take away benefits I mentioned, and the linked OpenAI post mentions they will address this.
aktuel 4 hours ago [-]
it's really good until it isn't and you can't tell the difference
GuinansEyebrows 4 hours ago [-]
> Why would I not tell AI about my personal stuff?
aside from my economic tilt against for-profit companies... precisely because your personal stuff is personal. you're depersonalizing by sharing this information with a machine that cannot even attempt to earnestly understand human psychology in good faith and then accepting its responses and incorporating them into your decision-making process.
> It's really good at giving advice.
no, it's not. it's capable of assembling words that are likely to appear near other words in a way that you can occasionally process yourself as a coherent thought. if you take it for granted that these responses constitute anything other than the mere appearance of literally the most average-possible advice, you're abdicating your own sense of self and self-preservation.
press releases aside, time and again these companies prove that they're not interested in the safety or well-being of their users. cui bono?
koakuma-chan 4 hours ago [-]
If these models give the most average possible advice, then average advice I get from humans must be around terrible. If you use Gemini, you can enable grounding and you will be able to see the source.
GuinansEyebrows 2 hours ago [-]
maybe so, but you also probably have the life-changing and highly-enriching opportunity to meet new people and develop meaningful relationships nearly every single day.
koakuma-chan 1 hours ago [-]
You're absolutely right.
astrange 4 hours ago [-]
> no, it's not. it's capable of assembling words that are likely to appear near other words in a way that you can occasionally process yourself as a coherent thought.
It doesn't emit words at all. It emits subword tokens. The fact that it can assemble words from them (let alone sentences) shows it's doing something you're not giving it credit for.
> literally the most average-possible advice
"Average" is clearly magical thinking here. The "average" text would be the letter 'e'. And the average response from a base model LLM isn't the answer to a question, it's another question.
GuinansEyebrows 2 hours ago [-]
i'm comfortable enough including the backend process of assembling strings that appear to be words in the general description of "assembling words".
re: average - that's at a character level, not the string level or the conceptual level that these tools essentially emulate. basically nobody would interpret "eeee ee eeeeee eee eeeeeeee eee ee" as any type of recognizable organized communication (besides dolphins).
vorpalhex 3 hours ago [-]
Am I depersonalizing by sharing my problems with my stuffed animal or my journal?
ELIZA has existed in emacs for a long, long time.
Humans are funny creatures who benefit frequently from explaining the problem slowly and having it fed back to them.
And for many, average advice really is a dramatic improvement over their baseline.
koakuma-chan 3 hours ago [-]
> Humans are funny creatures who benefit frequently from explaining the problem slowly and having it fed back to them.
Yeah, sometimes I realize the solution in the process of writing a GitHub issue.
GuinansEyebrows 2 hours ago [-]
> Am I depersonalizing by sharing my problems with my stuffed animal or my journal?
you're strengthening your personality with these activities. neither your journal nor your stuffed animal (cute :) ) respond to you with shallow recreations of thought - they allow you to process your internal thoughts and feelings in an alternative and self-reinforcing way.
> ELIZA has existed in emacs for a long, long time.
ELIZA doesn't really give advice, does it? it's a fun toy, and if there's any serious use for it, it's similar to journaling or rubber-ducking in that it's just getting you to talk about things yourself.
reaperducer 5 hours ago [-]
Why would I not tell AI about my personal stuff? It's really good at giving advice.
Define "good" in this context.
Being able to ape proper grammar and sentence structure does not mean the content is good or beneficial.
Chris2048 7 hours ago [-]
This is % though. Is that because the people that use it for work, are still using for work (or more even); because some have stopped using it for work, or because there is an influx of people using it for other things that never have, or will, use it for work.
xg15 4 hours ago [-]
Good thing crippling depression and suicidal ideation automatically stop when you turn 18...
everdrive 4 hours ago [-]
I was originally upset about AI age-identification, but I think this might be the least-bad option given the route we're on:
- clearly the wider public is moving towards REAL identification to be online. Anything which delays or prevents this is probably welcome.
- It's easy to game, but also easy to be misclassified. (this isn't a positive, but I think there's no avoiding this unless I have to provide my passport or driver's license or something)
It's not impossible to think that this could satisfy enough people to prevent the death of the anonymous internet.
rchaud 3 hours ago [-]
> clearly the wider public is moving towards REAL identification to be online.
No they're not. Nobody voted for that. It is simply being imposed on people via government mandates.
dcow 2 hours ago [-]
In most places people are the government. Nobody directly votes for laws. And you might be surprised if they did.
rchaud 19 minutes ago [-]
In America, winning the most votes doesn't guarantee the power to form national government. There is no limit on opaque corporate donations, which can be masked via PACs, while individual citizens have strict limits and face jail time for exceeding them.
And even if a majority vote was enough to form government, the so called will of the people can be overruled by an unelected judiciary with lifetime appointments. You're really stretching the definition of "people are the government".
kevin_thibedeau 3 hours ago [-]
It will be as effective as the Leisure Suit Larry age verification.
1970-01-01 3 hours ago [-]
>Some of our principles are in conflict
Sam is missing the forest for the trees. Conflicting principles is a permanent problem at the CEO level. You cannot 'fix' conflicting principles. You can only dress them up or down.
ddtaylor 7 hours ago [-]
I'm fairly certain all LLMs can do the basic sentiment analysis needed to render a response like "This is something you really need to talk to a professional about. I have contacted one that will be in this conversation shortly."
shmel 6 hours ago [-]
Yeah, right. Just one step from "Based on your comments about recent political events you are engaging into a thought crime. A police officer will join this conversation shortly".
rchaud 3 hours ago [-]
They're about as likely to disclose that as law enforcement would let someone know that a judge has signed a wiretap warrant for their phone.
7 hours ago [-]
bell-cot 7 hours ago [-]
Whether or not that's true - no CFO would want to pay for it, and no Chief Legal Officer would want to assume the risks.
raminyt 5 hours ago [-]
Until some Mr. President or somebody sits them in his stately room and tells them it is in their best interest to really rethink that and that there is really NO PROBLEM. This is not really meant as a joke.
BrawnyBadger53 6 hours ago [-]
It's interesting to see so many people convinced it's related to their specific media they saw (all unique from each other). I think this is more indicative that the issue is just well known and this is a response to the issue at large rather than a specific instance.
Sparkle-san 6 hours ago [-]
Having freshly heard the NY Times piece on a recent teen suicide stemming from ChatGPT, I don't think it's wrong to assume that it's playing a large role here as what ChatGPT did in this instance was egregious. Feel free to judge for yourself.
We have existing precedent that encouraging someone to kill themselves can result in you being criminally responsible. Is software doing its best to be human that different?
Yeah! That will show all those people with serious mental health problems!
DaSHacka 3 hours ago [-]
I mean, natural selection?
e40 6 hours ago [-]
Just today The Daily pod is about people who develop unhealthy relationships with ChatGPT. A teenage boy committed suicide and a good part of the episode is about that. As a parent, heartbreaking to listen to...
charcircuit 7 hours ago [-]
>We’re building an age-prediction system to estimate age based on how people use ChatGPT.
>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.
voakbasda 6 hours ago [-]
This is why one should never say anything sensitive to a cloud-hosted AI.
Local models and open source tooling are the only means of privacy.
SoftTalker 6 hours ago [-]
Same goes for doctors, therapists, lawyers, etc. then. They all ultimately have the responsibility to involve authorities if someone is expressing evidence of imminent harm to himself or others.
godshatter 4 hours ago [-]
Yep, I'll be using something like gpt4all and running things locally just so I don't get caught up in something by some online AI calling the authorities on me. I don't plan to talk about anything anyone would be concerned about, but I don't trust these things to get nuance.
mtlmtlmtlmtl 3 hours ago [-]
I have to say, when I see a post by a company like OpenAI about "safety, freedom and privacy", I can't keep a straight face. They might as well title the piece "If you don't mind, we'd like to gaslight you across several paragraphs". No thanks.
wagwang 6 hours ago [-]
For those who don't know, this is probably in response to the tucker carlson interview.
anon1395 7 hours ago [-]
This was probably made in response to that bad press from that ex-yahoo employee.
trallnag 8 hours ago [-]
Sorry, but what is the "over 18 years old" experience on ChatGPT supposed to be? I just tried out a few explicit prompts and all of them get basically blocked. I've been using it for quite some time now and have paid for it in the passed. So I should be recognized as a grown-up
enmyj 5 hours ago [-]
lol
bayindirh 8 hours ago [-]
TL;DR: We're afraid from what happened and ChatGPT probably screwed up badly in "that teen case". We're trying to do better, so please don't sue us this time.
TL;DR2: Regulations are written with blood.
d2049 8 hours ago [-]
Reminder that Sam Altman chose to rush the safety process for GPT-4o so that he could launch before Gemini, which then led directly to this teen's 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
Can you comment on your own opinions, or take-aways from those articles, rather than just link dump?
decremental 5 hours ago [-]
[dead]
Chris2048 7 hours ago [-]
It's be worse that the bot becomes a nannying presence - either pre-emptively denying anything negative based on the worst-case scenario, or otherwise taking in far more context than it should.
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.
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.
Or maybe, deep in the terms and conditions, it will add you to Altman's shitcoin company[0]
[0]https://en.wikipedia.org/wiki/World_(blockchain)
Oh brilliant. The same authorities around the world that regularly injure or kill the mentally ill? Or parents that might be abusing their child? What a wonderful initiative!
How long will it take for someone to accidentally SWAT themselves?
This current approach is a net negative, but the TLD idea actually makes sense to me.
Here's a thought experiment: you're a gay person living in a country where being gay is illegal and results in a death penalty. You use ChatGPT in a way which makes your sexuality apparent; should OpenAI be allowed to share this query with anyone? Should they be allowed to store it? What if it inadvertently leaks (which has happened before!), or their database gets hacked and dumped, and now the morality police of your country are combing through it looking for criminals like you?
Privacy is a fundamental right of every human being; I will gladly die on this hill.
There's a reason why e.g. banks want to have all critical systems on premises, under their physical control.
Why do people speak of ML/AI as an entity when it is a tool like a microwave oven? It is a tool designed to give answers, even wrong ones when the question is nonsensical.
[0] https://www.ala.org/advocacy/advleg/federallegislation/theus...
[1] https://www.ala.org/advocacy/intfreedom/statementspols/other...
The only secure position for a company (provided that the company is not interested in reading your communication) is the position of a blind carrier that cannot decrypt what you say; e.g. Mullvad VPN demonstrated that it works. I don't think that an LLM hosting company can use such an approach, so...
It depends. If you're speaking to a doctor or a lawyer, yes, by law they are bound to keep your conversation strictly confidential except in some very narrow circumstances.
But it goes beyond those two examples. If I have an NDA with the person I am speaking with on the other end of the line, yes I have the "right" to "force" the other person to keep our conversation private given that we have a contractual agreement to do so.
As far as OpenAI goes, I'm of the opinion that OpenAI - as well as most other businesses - have the right to set the terms by which they sell or offer services to the public. That means if they wanted a policy of "all chats are public" that would be within their right to impose as far as I'm concerned. It's their creation. Their business. I don't believe people are entitled to dictate terms to them, legal restrictions notwithstanding.
But in so far as they promise that chats are private, that becomes a contract at the time of transaction. If you give them money (consideration) with the impression that your chats with their LLM are private because they communicated that, then they are now contractually bound to honour the terms of that transaction. The terms that they subjected themselves to when either advertising their services or in the form of a EULA and/or TOS presented at the time of transaction.
When I'm talking to my doctor, or lawyer, or bank. When there's a signed NDA. And so on. There are circumstances where the other person can be (and is) obliged to maintain privacy.
One of those is interacting with an AI system where the terms of service guarantee privacy.
If I start any kind of company, I cannot just invent new rules for society via ToS; rather the society makes the laws. If we just make a simple law that states minors are not allowed to access the web and/or access any user generated content (including chat), it won't need to be enforced by every site/app owner, it would be up to the parents.
The same way schools cannot decide certain things for your children (even though they regularly over reach...).
We need better parenting. How about some mandatory parenting classes/licenses for new parents? Silly right? Well its just as silly as trying to police the entire internet. Ban the kids from internet and the problem will be 95% solved.
Apparently he came across as articulate enough that she couldn't tell the difference between his posts and that of any random adult spewing their political BS.
This predated ChatGPT so just imagine how much trouble a young troll could get up to with a bit of LLM word polishing.
20 years ago it was common for people to point out that the beautiful woman their friend was chatting up is probably some 40 year-old dude in his mom's basement. These days we should consider that the person making us angry in a post could be a bot or it could be some teenager just trying to stir shit up for the lulz.
Dead Internet theory might be not be literally true, but there's certainly a lot of noise vs signal.
my guy
It’s a big reason why tech stocks are still high IMO. It’s where today’s kids will spend their time on when they become old enough to spend their own money.
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
https://pca.st/episode/73690b66-8f84-4fec-8adf-e1a02d292085
Yay, proactive censorship?
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.
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
I just want to be explicit that my point isn't, "So what?" so much as, "We BEEN on that slippery slope." Social expectations (and related formal protocols in business) could do with some acknowledgement of our society's inherent... wait for it... ~diversity~.
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
aside from my economic tilt against for-profit companies... precisely because your personal stuff is personal. you're depersonalizing by sharing this information with a machine that cannot even attempt to earnestly understand human psychology in good faith and then accepting its responses and incorporating them into your decision-making process.
> It's really good at giving advice.
no, it's not. it's capable of assembling words that are likely to appear near other words in a way that you can occasionally process yourself as a coherent thought. if you take it for granted that these responses constitute anything other than the mere appearance of literally the most average-possible advice, you're abdicating your own sense of self and self-preservation.
press releases aside, time and again these companies prove that they're not interested in the safety or well-being of their users. cui bono?
It doesn't emit words at all. It emits subword tokens. The fact that it can assemble words from them (let alone sentences) shows it's doing something you're not giving it credit for.
> literally the most average-possible advice
"Average" is clearly magical thinking here. The "average" text would be the letter 'e'. And the average response from a base model LLM isn't the answer to a question, it's another question.
re: average - that's at a character level, not the string level or the conceptual level that these tools essentially emulate. basically nobody would interpret "eeee ee eeeeee eee eeeeeeee eee ee" as any type of recognizable organized communication (besides dolphins).
ELIZA has existed in emacs for a long, long time.
Humans are funny creatures who benefit frequently from explaining the problem slowly and having it fed back to them.
And for many, average advice really is a dramatic improvement over their baseline.
Yeah, sometimes I realize the solution in the process of writing a GitHub issue.
you're strengthening your personality with these activities. neither your journal nor your stuffed animal (cute :) ) respond to you with shallow recreations of thought - they allow you to process your internal thoughts and feelings in an alternative and self-reinforcing way.
> ELIZA has existed in emacs for a long, long time.
ELIZA doesn't really give advice, does it? it's a fun toy, and if there's any serious use for it, it's similar to journaling or rubber-ducking in that it's just getting you to talk about things yourself.
Define "good" in this context.
Being able to ape proper grammar and sentence structure does not mean the content is good or beneficial.
- clearly the wider public is moving towards REAL identification to be online. Anything which delays or prevents this is probably welcome.
- It's easy to game, but also easy to be misclassified. (this isn't a positive, but I think there's no avoiding this unless I have to provide my passport or driver's license or something)
It's not impossible to think that this could satisfy enough people to prevent the death of the anonymous internet.
No they're not. Nobody voted for that. It is simply being imposed on people via government mandates.
And even if a majority vote was enough to form government, the so called will of the people can be overruled by an unelected judiciary with lifetime appointments. You're really stretching the definition of "people are the government".
Sam is missing the forest for the trees. Conflicting principles is a permanent problem at the CEO level. You cannot 'fix' conflicting principles. You can only dress them up or down.
https://www.nytimes.com/2025/08/26/technology/chatgpt-openai...
https://en.wikipedia.org/wiki/Death_of_Conrad_Roy
>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.
TL;DR2: Regulations are written with blood.
https://news.ycombinator.com/item?id=45026886
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.