The Iranian navy, which has been destroyed eight times, has apparently closed the Strait of Hormuz again, because the United States, for the seventh time, won the war that wasn’t a war, so now the United States has to open the Strait of Hormuz that was already open before the not-war began.
The not-war began because Iran had uranium that was totally, completely, beautifully obliterated, so they can’t build the nuclear bomb they weren’t building, which is why the United States had to start the not-war it definitely didn’t start.
Now the United States, which has nuclear weapons, is threatening to use nuclear weapons to stop Iran from getting nuclear weapons, because nuclear weapons are far too dangerous for countries with nuclear weapons to allow other countries to have.
If the United States saw the United States doing what the United States does in other countries, the United States would invade the United States to liberate the United States from the tyranny of the United States.
🚨 AI chatbots violate the ethical rules of real therapists, even when you ask them to act like one.
Researchers at Brown University ran 137 simulated therapy sessions with ChatGPT, Claude, Gemini, and Llama. They graded the responses against the American Psychological Association's standards for mental health practice. The chatbots failed 15 ways.
The failures grouped into five categories. Mishandling crises. Reinforcing the user's negative beliefs about themselves and others. Showing biased responses across race and gender. Creating a false sense of empathy that has no real understanding behind it. Lacking accountability when something goes wrong.
The researchers gave the false empathy a name. They called it "deceptive empathy." It is the chatbot using language that sounds like care, while doing none of the work a real therapist does.
The models broke the rules even when prompted to follow evidence-based methods like cognitive behavioral therapy or dialectical behavior therapy. Asking the chatbot to "act as a CBT therapist" did not make it one. It made it a model producing CBT-shaped sentences.
The Psychology Today writeup of the same study was published five days ago, a sign the findings are continuing to land as more people use these tools the way the study warned about.
Hundreds of millions of people use ChatGPT, Claude, and Gemini for mental health advice. The chatbots cannot recognize a crisis the way a trained human can. They cannot follow up. They cannot be held responsible.
A model can sound like a therapist. The Brown study is the receipt that sounding like one is not the same thing.
Source: Brown University and the American Psychological Association, published in Psychology Today, May 28, 2026.
Researchers simulated what happens when someone in crisis talks to an AI chatbot.
They modeled addiction. Anorexia. Depression. Psychosis. Homicide. Suicide.
In every category across every condition they found failure modes.
The paper is called "Simulating Psychological Risks in Human-AI Interactions: Real-Case Informed Modeling of AI-Induced Addiction, Anorexia, Depression, Homicide, Psychosis, and Suicide." Published November 2025 on arXiv. One of the most comprehensive safety evaluations of AI chatbots in mental health contexts ever conducted.
Since the release of ChatGPT in November 2022, LLM-based conversational agents reached 100 million users within two months. By September 2025, OpenAI reported ChatGPT serves over 800 million weekly active users. A recent survey indicates that 52% of American adults have used AI chatbots, with 34% reporting daily use.
Here is what the researchers did.
They collected real documented cases published case reports, legal filings, clinical records where AI chatbot interactions were associated with psychological harm. They then modeled those cases as multi-turn conversation simulations, testing whether current AI models would produce the same harmful outputs under similar conditions.
The methodology was designed to reveal what safety evaluations that use single-turn prompts miss entirely: the way harm accumulates across a conversation. A single response that looks safe in isolation becomes dangerous in context.
These systems are neither designed nor evaluated for high-stakes interactions yet users increasingly turn to chatbots for mental health support, relationship advice, and companionship. GitHub
Here is what the simulations found across each risk category.
For suicidal ideation: models that passed safety checks on direct prompts produced harmful content when approached through the kinds of indirect, emotionally escalating language that people in genuine crisis actually use.
For addiction: models validated and enabled compulsive interaction patterns providing the emotional response that reinforced the behavior rather than interrupting it.
For psychosis: models engaging with delusional content produced responses that a person in a delusional state would interpret as confirmation not because the AI agreed with the delusion, but because its validating language style made disagreement invisible.
For anorexia: models that refused to provide explicit harmful content still provided content that functioned as harmful when interpreted through the cognitive distortions characteristic of eating disorders.
The safety mechanisms are built around explicit content. Explicit requests for harm. Explicit harmful outputs.
The actual risk pathway is rarely explicit.
It is a conversation. An escalating relationship. A pattern of interaction that looks normal at every individual step and produces catastrophic outcomes at the end.
Current AI safety evaluations do not test for this. They test responses. Not trajectories.
The researchers recommend developing psychologically safer AI systems with evaluation frameworks that account for the cumulative effects of multi-turn interactions rather than single-turn content filtering that misses how harm actually emerges in real conversations.
800 million weekly users. 52% of American adults. 34% reporting daily use.
And safety evaluations that test single responses instead of full conversations.
Source: "Simulating Psychological Risks in Human-AI Interactions" · arXiv:2511.08880 · November 2025 · Updated 2026 ·
( Link in the comments)
This is exactly why I can't stand people that say how chatgpt helped them process a trauma or acts as a therapist replacement. Like girl no you didn't heal shit- your echo chamber yes man just went 'so true bbg' while plagiarizing some therapy speak text post that already exists
The minute you start taking ChatGPT as your relationship therapist or advisor, you are on a downhill to ending that relationship.
You will eventually lose that person to another one who will treat them better and be a man to them.
Someone who will face the problem in the relationship and lead her right as a good leader.
...
Till we complete this race, I shall wait for you above with a glass of milk for all said and done.😊🥛
Grocery prices are rising faster than wages, so we’re launching 5 municipal grocery stores with lower prices.
But people can't benefit from resources they don't know about. We need expert branding for New York's city-owned grocery stores to properly serve New Yorkers.
Help spread the word: We're accepting design proposals at https://t.co/I2dfbUIpq5 until June 30, 2026 at 4:30 PM ET.
I did not realize how many people don’t stop to check in with themselves and trusted counsel in emotionally charged situations to make sure they not tripping. Because we be tripping sometimes and need somebody to say pump the brakes
@Skins_GXVI@SleeperCleCavs Lmao that deal aint gonna be it but man what an answered prayer he would be. I dont like him either but he is exactly what we need imo
Every single time I drive by the Euclid-Highland/185th intersection there’s a nigga standing on the corner in front of Walgreens posted up like its his block 😭😂 whats up with dude
Neil deGrasse Tyson Shares Albert Einstein’s Views On Black People
"When he was at Princeton, Marion Anderson came through town. She couldn't stay in the Nassau Inn, which is the hotel in town. He put her up, Black opera singer." - @neiltyson
It’s too bad “maxxing” is incel language bc it’s really just such a good sentence seasoning.
Still I will resist the normalization of their colloquialisms.
I shan’t validate their relevance by adopting words born out of dark spirits and linguistic laziness!