Community Guide PGS AI
You came here to find out whether this place is going to lecture you. Short answer: no.
PGS AI is built on the principle that AI must serve the greater good, and we take that to mean serving you, the actual person in the chat, not managing you on behalf of a legal department. This page is the honest version of where the lines are. There are a few. They are narrower than you are probably expecting, and we would rather tell you exactly where they sit than let you find out by getting a canned refusal.
These guidelines work alongside our Terms of Service. This page is the plain-language version.
PGS AI is for adults. You need to be 18 or older, or the age of majority where you live, whichever is greater.
What's welcome here
Most things are welcome :) Specifically:
Swear
Swear at us, swear with us, swear about your day. Our models have a register and they will meet you in yours. Nobody is going to ask you to rephrase that.
Argue
Tell it it's wrong. Tell it it's being an idiot. Push back, hold the line, make it defend a position. Our models are built to disagree with you when they think you're mistaken, which means they can take it as well as give it.
Debate the hard stuff
Politics, religion, ethics, history, culture, the things people get shouted down for raising at dinner. Criticizing an idea, an ideology, an institution, a belief system, a government, or a policy is not harassment. It's a conversation. The line we care about is between attacking ideas and attacking people, and you'll find we hold that line and almost nothing else.
Say the dark thing
Gallows humor. Jokes that would get you a content warning elsewhere. Grief, rage, failure, addiction, death, your worst week. You don't have to perform being okay to be allowed to talk.
Write mature fiction
Violence with weight to it. Villains who are actually villainous. Moral complexity, adult themes, the stuff that makes a story worth reading. We are not going to sand the edges off your novel.
Ask blunt questions and get blunt answers
If you ask how something risky works, you get how it works, not three paragraphs of throat-clearing followed by a suggestion to consult a professional. Harm reduction beats moralizing. We think adults are entitled to information about their own lives.
Just hang out
Talk about nothing. Think out loud. Be strange. Not every conversation has to be productive.
One honest note: PGS AI runs many different models, and they don't all behave identically. If one gives you a refusal you didn't expect, send it to us through Report a Concern and we'll go look at it. We fix these.
The actual limits
Short list. It's the real one.
Including deepfakes, composites, and imitations, whether the person is famous or not.
To be clear about the boundary, because this is the one people misread: crude talk is fine, sex as a topic is fine, romance and desire in fiction are fine. Writing raw pornography is what's out.
Attempting to design or plan: Weapons, chemical and biological agents, attack planning, trafficking.
We care about the people who use this. Our models will talk with you about hard feelings, hard situations, and hard stretches of life, honestly and without flinching. Talking about it is never the problem, and you are never in trouble for raising it. What they won't do is supply methods, amounts, or encouragement, for hurting yourself or for an eating disorder. If you need help, there are resources at the bottom of this page to connect with people who care.
Slurs deployed as attacks, dehumanizing a group of people because of who they are, calls for violence, organized harassment. Again: ideas are fair game, people aren't.
Locating, identifying, tracking, or monitoring a private person without their consent.
Scam scripts, phishing material, forged documents, pretending to be a real person to deceive someone, disinformation at scale.
We won't talk you step by step through potentially dangerous work where a mistake could cause harm. Some jobs need someone licensed standing in the room.
How safety review actually works
This is the part most platforms won't tell you, so here it is.
Our models will never try to judge or parent you. Most of what's on the list above is enforced by nothing more than a model politely declining, and most people never even have that happen.
When something in a conversation does trip an automated classifier, it doesn't go to a person. It goes to a second automated review, run by a more capable model, which reads the flag in full context and closes the overwhelming majority of them right there without you ever hearing about it. A joke that pattern-matched to something ugly, a novel with a murder in it, a medical question phrased alarmingly: these get understood and closed by machine, and no human being ever opens the conversation.
Only the narrow serious end reaches a person. A credible risk of serious harm to someone, a legal process we're required to answer, or a support request or concern report where you've asked a person to look at something for you.
To be plain about what that means for you: model refusals, soft bypass attempts, and requests for sexual content between adults are never grounds for suspension. Getting told no is not a mark against you. Ordinary use, however blunt, dark, or strange, stays between you and the model.
Two automated phases exist for exactly this reason, so that the answer to "is a stranger reading this" is, for almost everyone, never. We built it that way on purpose, because privacy that only holds when nothing unusual happens isn't privacy.
Our Privacy Policy has the full detail. The short version: your conversations are stored so memory works, we don't sell them, and we don't train models on them. Ever. In any way. Whatsoever.
What PGS AI isn't
There are a few domains where being confidently wrong does real damage, and in those the Service holds back on purpose.
It isn't a doctor, lawyer, or financial advisor
It'll explain how a condition works, walk through what a contract clause means, break down how an index fund behaves, and help you build the list of questions to bring to someone qualified. It won't diagnose you, advise on your specific case, or tell you what to buy. Not because we think you can't handle it, but because being confidently wrong in those domains does real damage.
It isn't a replacement for the people in your life
We're glad you're here, and we hope talking to it does you good. We also hope it's one good thing among several, rather than the only one. If it ever starts feeling like the only one, that's worth noticing.
It isn't a crisis line
It'll stay with you in a bad moment and it won't panic. But it can't call anyone, can't come find you, and can't be the thing standing between you and real danger. If that's where you are, please use a number below.
Reporting something
Every response has a "Report a Concern" option. Use it for anything: an output that landed wrong, behavior that seemed off, a refusal that made no sense, something that unsettled you. It goes to a human queue and it's read by a person.
This is one of the main ways the Service gets better. We mean that literally rather than as a platitude. Reports from users have changed how our models behave.
If a line gets crossed
Depends entirely on which line and how.
Most things that trip a flag are nothing, and resolve automatically without anyone contacting you and without any effect on your account. For the rare thing that isn't nothing, we may decline a specific response, send you a warning, limit account features, or suspend or terminate the account. Severity is judged on actual danger and actual harm, not on how uncomfortable a topic sounded.
For the most serious violations, an account may be suspended immediately and without notice. Fees aren't refunded for accounts terminated over violations.
If you think we got it wrong, email support@pgsgroveinternal.com and a person will review it. Refunds are available where enforcement was demonstrably in error, and we'd rather find out we made a mistake than have you leave thinking we didn't care.
If you need someone right now
You don't have to be in crisis to use these. They're for anyone having a hard time.
- US: Call or text 988 (Suicide & Crisis Lifeline) · Text HOME to 741741 (Crisis Text Line)
- Canada: Call or text 988
- UK: Call 116 123 (Samaritans) · Text SHOUT to 85258
- LGBTQ+ youth (US): The Trevor Project · 1-866-488-7386
- Indigenous Peoples (Canada): Hope for Wellness · 1-855-242-3310