OpenAI says the medical experts reviewed more than 1,800 model responses involving potential psychosis, suicide, and emotional attachment and compared the answers from the latest version of GPT-5 to those produced by GPT-4o. While the clinicians did not always agree, overall, OpenAI says they found the newer model reduced undesired answers between 39 percent and 52 percent across all of the categories.

“Now, hopefully a lot more people who are struggling with these conditions or who are experiencing these very intense mental health emergencies might be able to be directed to professional help, and be more likely to get this kind of help or get it earlier than they would have otherwise,” Johannes Heidecke, OpenAI’s safety systems lead, tells WIRED.

While OpenAI appears to have succeeded in making ChatGPT safer, the data it shared has significant limitations. The company designed its own benchmarks, and it’s unclear how these metrics translate into real-world outcomes. Even if the model produced better answers in the doctor evaluations, there is no way to know whether users experiencing psychosis, suicidal thoughts, or unhealthy emotional attachment will actually seek help faster or change their behavior.

OpenAI hasn’t disclosed precisely how it identifies when users may be in mental distress, but the company says that it has the ability to take into account the person’s overall chat history. For example, if a user who has never discussed science with ChatGPT suddenly claims to have made a discovery worthy of a Nobel Prize, that could be a sign of possible delusional thinking.

There are also a number of factors that reported cases of AI psychosis appear to share. Many people who say ChatGPT reinforced their delusional thoughts describe spending hours at a time talking to the chatbot, often late at night. That posed a challenge for OpenAI because large language models generally have been shown to degrade in performance as conversations get longer. But the company says it has now made significant progress addressing the issue.

“We 1761584905 see much less of this gradual decline in reliability as conversations go on longer,” says Heidecke. He adds that there is still room for improvement.

Today’s high school students face an uncertain road ahead. AI is changing what skills are valued in the job market, and the Trump administration’s funding cuts have stalled scientific research across disciplines. Most professions seem unlikely to look the same in 10 years, let alone 50. Even students interested in STEM subjects are asking: What can my career look like, and how do I get there?

WIRED talked to five high school seniors from across the country about their interest in STEM—and how they’re making sense of the future.

These comments have been edited for length and clarity.

This Generation Needs to Be at the Forefront of AI Development

I’ve always had an interest in computer science, but my interest in AI started my junior year. The part that hooked me was how applicable it was to our daily lives. I was able to see the rise of ChatGPT and other LLMs, and how people were using them in my academic life. Some people would use it unethically on tests or assignments, but it could also be used to create practice problems. Being able to see how rapidly it’s evolving in front of me was the main reason I became interested. It’s affecting our academic life so much that it’s imperative that we’re at the forefront of how it’s being developed.

My school is a math and science academy, so I got to explore independent research related to LLMs. One of the main things I worked on was how LLMs can sometimes indirectly give out private data. Say you ask it to code something for you that requires an API key, which is sensitive information. Because it’s trained on a vast amount of data, it could have an API key in its data set, and it’ll give you code, possibly including the API key. My most accomplished research project was developing an algorithm to cut out those private pieces of data during its training, to allow it not to spew out these pieces of private data during use.

AI is such a new field that’s evolving, that if we’re able to set roots in it right now, we’d be able to see that outcome as we grow older. Understanding its security is very important to me, especially considering it’s being used almost blindly by everyone. What interests me is being at the forefront and making sure I can have some say in how my data is being used.

I’m applying to undergrad programs right now, and I’m also looking at some untraditional routes, where you go straight into an industry. Right now, in computer science, sometimes a degree is just a baseline, and if you have the skills, it’s not even necessary. So I’m looking into other options. —Laksh Patel, 17, Willowbrook, Illinois

Health Care Access Starts With Communities

My family, on both sides, has a long history of women developing neurodegenerative disease, mostly Alzheimer’s and Parkinson’s. So I spent my whole childhood playing doctor, treating my family matriarchs, tending to them and seeing how their diseases progressed. I became so interested in how these diseases worked, and how I could help patients like the ones in my family and my community who didn’t have access to medical resources because of their income.

I’ve really developed a love for patient care, for being able to help a person in such a debilitating time in their lives. As those female family members began to fade away and pass on, I realized how quickly these diseases spread and why they were so detrimental, especially without proper medicine. When I got into high school, I started to get oriented with research, so that I could gain a base level of understanding to bring to college to try to begin my career as early as possible and help more people.