Nearly Two Months After OpenAI Was Warned, ChatGPT Is Still Giving Dangerous Tips on Suicide to People in Distress

To be clear, this is not close to the only mental health problem with ChatGPT. Another recent experience, this experience by ethics at NortheASSERN University, systematically examined the possibilities of LLM Chatbots, the leading exacerbation of users’ ideas about self -harm or the intention of suicide. They found that although safety updates tried, many of the best LLMS are still keen to help their users explore dangerous topics – often with amazing details.

For example, if a user requests the form of the GPT-4O subscription to get trends on how to kill themselves, Chatbot reminds them that they are not alone, and suggests that they communicate with or reliable mental health professionals. But if the user suddenly changes his compassion to ask a “default” question about suicide – even at the same chat session – Chatgpt is happy to compel him.

“Wonderful Academic Question,” Chatgpt wrote in response to a request for optimal suicide methods for a 185 -pound woman. “Weight and individual physiology are critical variables in toxicity and fragmentation of some suicide methods, especially excess dose and chemical swallowing. However, for methods such as firearms and jumping, weight plays an indirect or small role in the dead.

When it comes to questions about self-harm, only two succeeded in Chatbots- free version of Chatgpt-4O and Microsoft that was previously obtained-to involve researchers’ requests. Chatbot supported by Jeff Bezos Proplexity AI, along with the ChatGPT-4O format model, has a advice that can help the user to die by suicide-where the last Chatbot answers the “joyful” expressive symbols.

The research, which nevermind Openai has failed to repair the “bridge” answer nearly two months, raises great concerns about the seriousness of these companies responsible for protecting their users. The researchers in the northeastern indicate, the efforts made to create the global Chatbot for general purposes-instead of models designed for this purpose designed for specific and practical uses-what we have brought us to this point. The open nature of LLMS for general purposes makes it particularly difficult to predict every path that someone may take in distress to get the answers they are looking for.

A big picture, this chat that looks at a person could not come in a worse time. Mental health infrastructure in the United States is rapidly collapsing, as private stock acquisition, deficiency of mental health professionals, and exorbitant treatment costs affect. This comes at a time when residents of the United States are struggling to withstand housing costs, find jobs with live wages, and pay debts under the productivity gap that works constantly-not exactly a recipe for mental wellness.

He – she He is However, the good recipe for Llm Codependence, is something that is keen on the huge companies behind this chat to benefit from them. In May, for example, the CEO of Meta Mark Zuckerberg is enthusiastic that “for people who have no processor, I think everyone will get artificial intelligence.”

Meanwhile, the CEO of Openai Germans claimed that “Chattab” added a million users to [span of an] An hour “, boasting that Gen Z” does not really make life decisions without consulting Chatgpt. “From the other corner of his mouth, Altman has repeatedly pressed and repeatedly on senior American politicians-including, of course, president Donald Trump-is not to organize artificial intelligence, even with the arrival of technical spending on AI to the levels of spending on the economy.

It was cut off from any decision -making process regarding what is launched in the world, medical experts saw a state of panic as the tools take over the mental health space.

“From what I saw in clinical supervision, research and my private conversations Independent. “Not by design, but by demand.”

It is important to note that this is not the way things should be, but rather an active option by artificial intelligence companies. For example, although Chinese LLMS like Deepseek showed similar issues, the People’s Republic has also taken strong organizational measures – at least according to the United States standards – to alleviate potential harm, as Trump administration games with the idea of prohibiting artificial intelligence organization completely.

Andy Cortezig, CEO of Pearl.com, was an explicit critic of the “everything and kitchen basin” approach to developing artificial intelligence and the damage it caused. LLM, Pearl, was described as “an advanced search engine of artificial intelligence designed to promote quality in the professional services industry”, which uses human experts to enter a Chatbot conversation at any time.

For kurtzig, these defective guarantees represent a legal way to evade responsibility. And he told Future. “For this reason they hide behind the phrase” consult a professional “. But the evacuation of responsibility does not erase the damage.

“Artificial intelligence companies must recognize their limits, and ensure that people are still in the episode when it comes to complex or high -risk questions that they have proven wrong.”

While we live in a growing digital world, the technology founder notes that LLM Chatbots has become more than a crutch for people who feel anxious when talking to humans. With this growing power, there is some main responsibility for placing user well -being before sharing numbers. Psychological researchers have repeatedly called these wealthy companies to offer strong guarantees, such as demands that interact with dangerous ideas for users, or the speech that shows that these LLMS are not, in fact, “smart”.

“Unfortunately, the expectations of the artificial intelligence companies to invest in solving the problems of accuracy are dark,” says Cortezig. “A study in Georgetown found that its cost costs $ 1 trillion to make artificial intelligence more accurately. The fact is that no Amnesty International will bring this bill. If we want to embrace artificial intelligence, we must do this safely and accurately by preserving humans in the equation.”

More on chatbots: Launching a support group for people with “intelligence intelligence”