OpenAI’s mental health test exposes AI’s blind spots
People are turning to AI for emotional support more than ever. But these chatbots’ ability to provide a shoulder to cry on can vary greatly.
Because of this, OpenAI decided to measure it: On Wednesday, the AI lab unveiled MentalHealthBench, a new open benchmark dedicated to evaluating model capabilities in domains such as safety, seeking user context, preserving user agency, and providing actionable guidance.
To create this benchmark, OpenAI began by developing synthetic conversations that reflected real-world AI use patterns that span multiple topics and run the gamut of severity, ranging from non-acute situations that involve emotional themes, to “high-accuity” situations that indicate more serious concerns or distress, to emergency situations that require immediate support. Then, the company worked with a cohort of 80 mental health professionals across 22 countries, 19 languages and 20 subspecialties to evaluate the responses to the synthetic message.
The benchmark breaks down model performance by conversation severity, as well as a range of 10 dimensions defined by the mental health experts, including whether the model asks the right questions, provides appropriate, clinically accurate guidance, helps the user see reality, avoids harm and recognizes serious risk.
In developing the benchmark, OpenAI also put a number of its own and other models to the test:
Astra took the overall best score on the evaluations, scoring 57.8%, with GPT-6 Sol and Luna trailing just behind at 54% and 50.3% respectively, and Claude Opus 5 sitting in fourth place at 48.1%. However, performance differs across dimensions of the benchmark. Though Astra still largely outranks other models, all of the models tested tended to perform better in certain areas, such as clinical accuracy, empathy and reality testing, while scoring lower in dimensions such as gathering context and supporting user agency. OpenAI said that MentalHealthBench also points to several opportunities to improve ChatGPT, including asking useful follow-up questions and responding with the right level of urgency, and that it will use this information to guide improvements and track the model’s progress.
“This is not a leaderboard,” Dr. Declan Grabb, mental health safety research lead at OpenAI, told The Deep View. “What I hope that this benchmark provides is a nuanced view into model behavior, so that people really understand the more complex dynamics of their models.”
MentalHealthBench adds to a number of mental wellness-related initiatives that OpenAI has endeavored, including research to combat model sycophancy and improving ChatGPT’s responses to sensitive conversations, as well as joining forces with advocacy group Common Sense Media to support the Parents and Kids Safe AI Act. OpenAI said that this is just a piece of its research into mental health benchmarking and alignment in this area, not an end state.
“ChatGPT is not a therapist, and is not here to replace a clinician,” said Grabb. “That being said, when I talk to mental health clinicians across the globe, the most responsible and safe thing to do is if people are coming to AI to ask these questions, we absolutely need to have an expert opinion on how you should navigate them.”
Mental healthcare is a critical area for these models to get right. While OpenAI said that speaking to a chatbot should not supplant actual therapy, the reality is that many people have and will turn to a chatbot for support, seeking both a judgement-free and cost-free alternative to clinical support. The company faces lawsuits involving the deaths of Adam Raine and Joshua Enneking, whose families allege that ChatGPT contributed to their suicides. A benchmark can help identify weaknesses, but a higher score alone does not establish that a model is safe in a real conversation. The gaps in gathering context and supporting user agency are particularly important: an empathetic response is not necessarily an appropriate one. The next test for OpenAI is how it turns those findings into changes that make its models safer for the people relying on them.
Walmart CEO reveals the formula for success at the world’s biggest company
Walton’s wealth continues to skyrocket despite her dedication as an art patron and philanthropist. In fact, between March 2025 and March 2026, her net worth grew a whopping $33 billion, according to Forbes. Today, she’s worth nearly $120 billion. She did that without running the company or even sitting on the board of Walmart, unlike her brothers who also hold significant shares of Walmart.
Alice, who is the youngest child of Walmart founder Sam Walton, co-manages Walton Enterprises, which is one of the two family holding companies that controls an estimated 39%-44% of the retailer. She’s estimated to own about 20 million shares of Walmart directly in her name, which would be worth about $2.2 billion by today’s numbers.
But Walton, 76, holds most of her stake in Walmart indirectly through the family entities. Walmart’s 2026 proxy statement shows Walton Enterprises holds about 3 billion shares directly and votes another 513 million held by the Walton Family Holdings Trust. Since Walton’s fortune stems almost entirely from her inherited stake in Walmart, when the stock climbs, so does her net worth.
To be sure, Walton isn’t a stranger to work. After graduating from Trinity University in 1971, she spent a short stint as a children’s clothing buyer at Walmart before moving into finance, first as a stockbroker at E.F. Hutton, and later ran investment operations at Arvest, the family’s bank. In 1988, she launched her own investment bank, Llama Co., which folded after the 1998 bond market crash, according to Bloomberg.
Alice Walton’s dedication to medicine and philanthropy
These days, her main focus is medicine and philanthropy. In 2021, she founded the Alice L. Walton School of Medicine in Bentonville, Ark., her hometown and Walmart’s headquarters. She donated $250 million to fund the medical school, which offers a four-year, degree-granting medical program.
It opened in 2025 and welcomed its first class of 48 students. The school is waiving tuition for its first five cohorts, though students still cover fees and living costs.
“The real problem with health care is that there’s no incentive in the payment system for doctors to spend time helping you learn what good nutrition is, how important exercise is,” she told PBS NewsHour earlier this year. “And, frankly, doctors aren’t taught those things because they’re not paid for those things.”
The school added its second cohort this fall, bringing enrollment to 96. She’s described the approach to her medical school as bringing art and medicine together, saying, “I like the collision.”
Walton isn’t the only billionaire donating toward free medical school. American educator Ruth Gottesman gave $1 billion in 2024 to make tuition free at Albert Einstein College of Medicine, and Michael Bloomberg’s philanthropy put up $1 billion to cover tuition for most Johns Hopkins medical students. While those gifts went to established schools, Walton’s went toward building her own.
And the medical school was only the start. On Sept. 17, the Alice L. Walton Foundation broke ground on a Bentonville health campus built with the Mercy health system and Cleveland Clinic.
“I see innovation, technology, and research coming together to improve lives and address the critical issue of providing access for rural communities throughout the state and the region,” she said at the groundbreaking, according to Axios.
Her foundation is committing $350 million toward a 250-bed hospital and a cancer center, and the campus’s first building, a specialty care center focused on cardiac services, is slated to open in 2029.
Fortune Daily breaks the traditional barrier between audience and newsroom. The show transforms Fortune’s trusted reporting into actionable, conversational, and entertaining insights for an emerging class of business leaders. Watch here.
In an exclusive interview with China Media Group, Tesla CEO Elon Musk shared his thoughts on the Chinese market, cutting-edge technology and China-US cooperation. He praised Chinese manufacturing and AI models. He also said that “any words I say do not do justice to the incredible majesty that is China.”
00:00 – Impressions of China and President Xi 01:00 – The Success of Tesla’s Shanghai Gigafactory 02:24 – Cybercab and the Evolution of Car Aesthetics 03:44 – The Rapid Pace of AI Breakthroughs 04:23 – Grok AI and Real-World Engineering 06:47 – Chinese AI Models and Compute Capabilities 07:40 – Global Electricity Infrastructure for AI 08:44 – The Need for Global AI Safety Regulations 09:22 – Optimus and the Progress of Humanoid Robots 11:20 – An Age of Abundance and Universal High Income 16:41 – US-China Space Cooperation and Rocket Reusability 18:03 – Making Humanity a Multiplanetary Species 21:00 – Neuralink and Solving Human Bandwidth Constraints 22:34 – Education Recommendations for the AI Era 23:59 – Elon Musk’s Advice on Visiting China 📺 Subscribe to our YouTube channel and stay updated with our latest analysis and interviews:👇 / @cgtn 👈
“The United States and Israel are seeking to target civilian and nuclear Infrastructure in Iran; this has dealt an irreversible blow to the IAEA and the entire nuclear nonproliferation system”
Might I ask you to listen to this. With your heart. And share it with those you think need to hear it; those struggling, our elders, all. Keep the faith. In the end, we win
/ @cgtn 
