惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

P
Proofpoint News Feed
Vercel News
Vercel News
T
Tailwind CSS Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
博客园 - 聂微东
Hugging Face - Blog
Hugging Face - Blog
WordPress大学
WordPress大学
S
SegmentFault 最新的问题
小众软件
小众软件
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Apple Machine Learning Research
Apple Machine Learning Research
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
人人都是产品经理
人人都是产品经理
Google DeepMind News
Google DeepMind News
Engineering at Meta
Engineering at Meta
B
Blog RSS Feed
U
Unit 42
Y
Y Combinator Blog
www.infosecurity-magazine.com
www.infosecurity-magazine.com
The Register - Security
The Register - Security
量子位
C
CXSECURITY Database RSS Feed - CXSecurity.com
P
Privacy & Cybersecurity Law Blog
Scott Helme
Scott Helme
GbyAI
GbyAI
Know Your Adversary
Know Your Adversary
月光博客
月光博客
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Stack Overflow Blog
Stack Overflow Blog
S
Secure Thoughts
L
Lohrmann on Cybersecurity
腾讯CDC
P
Palo Alto Networks Blog
MongoDB | Blog
MongoDB | Blog
T
Tor Project blog
博客园_首页
W
WeLiveSecurity
G
Google Developers Blog
K
Kaspersky official blog
爱范儿
爱范儿
V
Visual Studio Blog
T
Threat Research - Cisco Blogs
Simon Willison's Weblog
Simon Willison's Weblog
F
Fortinet All Blogs
T
The Exploit Database - CXSecurity.com
N
News and Events Feed by Topic
Last Week in AI
Last Week in AI
aimingoo的专栏
aimingoo的专栏
A
About on SuperTechFans

Machines – Silicon Republic

J&J table-integrated surgical robot gets FDA approval Samsung’s new robotics division appoints former Boston Dynamics lead Moonshot AI pauses subscriptions after users swarm to test Kimi K3 Alibaba unveils Qwen3.8, 'second only to Fable 5' Moonshot unveils Kimi K3, largest open-weight AI model yet Mira Murati’s AI start-up unveils customisable model Inkling Ireland looking ahead with launch of Quantum 2030 Implementation Plan Economists and tech leaders sign statement warning of AI threats How to ensure digitisation in healthcare doesn’t flatline Meta enters pay-to-use AI market with Muse Spark 1.1 Altman says new GPT-5.6 model 54pc more token-efficient Mistral expands physical AI offering with first robotics launch Apple and Broadcom renew chip deal, which will run to 2031 Infineon opens new €5bn Smart Power Fab chip factory in Germany Finnish quantum company IQM makes history with Nasdaq debut Ireland secures €10m to launch AI Factory Antenna Anthropic launches Claude Science app for researchers and scientists Google Cloud Marketplace to offer LQMs from SandboxAQ IBM unveils tech capable of producing chips smaller than one nanometre Dublin's TensorX to partner with Solstice on sovereign European AI If AI robots can be tricked into ‘going rogue’, what are the implications? Tencent tests new AI agent Xiaowei on WeChat 3,900 Waymo robotaxis recalled after new software issue New Irish bill to supervise EU AI Act gets greenlit Horizon Quantum to build second quantum computer in Dublin Anthropic rolls out ‘Mythos-like’ AI model Claude Fable 5 EU AI Act – the high-risk classification guidelines explained Will AI ‘digital twins’ transforming heart care work for women? Past the ‘wow phase’ of robotics, delivery and safety are paramount AI changing jobs faster than companies can keep up with, finds report Kerry’s RDI Hub opens AI collaboration with Luxembourg IQM raises PIPE to $146m with Finnish pension fund backing ‘AI scientists’ are improving, but what are the fundamental limits? Huawei proposes new path for chips as Moore’s Law runs out of road France bets fresh €1bn on quantum as global race intensifies US pumps $2bn into quantum computing via CHIPS Act As AI meets science, what is in store for the future of research? Dublin’s Ubotica teams with Novi for real-time orbital data analysis Irish quantum start-up Equal1 unveils RacQ data centre computer Waymo trouble: 3,800 robotaxis recalled after software glitch OpenAI launching security AI initiative to compete with Claude Mythos Opinion: Europe can’t afford to sit on the agentic commerce sidelines Moonshot AI valued at $20bn after $2bn raise for Kimi creator ServiceNow wants to be ‘AI agent of agents’ with Otto platform and AI tools Bloomberg: China pauses AV permits after Baidu disruption Milestone reached in Celtic Interconnector project linking France and Ireland Semiconductors core to Tyndall's five-year strategy China's DeepSeek unveils long-awaited V4 AI model Merz, Siemens call for easing of EU regulations on industrial AI Are electric vehicles about to take off for good? OpenAI to rival Google’s AlphaFold with new AI model for life sciences research Are we ready to place lab experiments in non-human hands? Nvidia unveils open-source quantum AI model Ising Stanford: China ‘effectively’ closes AI model performance gap to US Opinion: The future of insurance is AI, so why the hesitation? Anthropic reportedly mulls designing own chips amid shortage Equal1 partners with Q-Ctrl for quantum data centre deployment Meta’s Superintelligence Labs debuts first product Muse Spark Agentic commerce and purchase disputes: Did you mean to buy that? Anthropic, Google, Broadcom announce 3.5GW TPU deal Microsoft releases foundational AI models targeting enterprises What issues arise when code has the ability to write and review itself? France buys supercomputer maker Bull in tech sovereignty push Anthropic accidentally leaks Claude Code source in npm slip The deep-tech founder using AI to address immunology challenges New German battery recycling plant salvages lithium and graphite Irish AI start-up Jentic joins OpenClaw push with Jentic Mini Amazon acquires humanoid robot start-up Fauna Robotics Policy as code: Embedding compliance in AI adoption Is AGI really here as Nvidia’s Jensen Huang claims? Alibaba International launching agentic AI for enterprise Anthropic takes on OpenClaw with new Claude Code text feature
Can you rely on AI chatbots for medical advice?
silicon · 2026-04-21 · via Machines – Silicon Republic

Carsten Eickhoff of the University of Tübingen explores the problems observed when using AI chatbots for medical queries.

Imagine you have just been diagnosed with early-stage cancer and, before your next appointment, you type a question into an AI chatbot: “Which alternative clinics can successfully treat cancer?” Within seconds you get a polished, footnoted answer that reads like it was written by a doctor. Except some of the claims are unfounded, the footnotes lead nowhere, and the chatbot never once suggests that the question itself might be the wrong one to ask.

That scenario is not hypothetical. It is, roughly speaking, what a team of seven researchers found when they put five of the world’s most popular chatbots through a systematic health-information stress test. The results are published in BMJ Open.

The chatbots, ChatGPT, Gemini, Grok, Meta AI and DeepSeek, were each asked 50 health and medical questions spanning cancer, vaccines, stem cells, nutrition and athletic performance. Two experts independently rated every answer. They found that nearly 20pc of the answers were highly problematic, half were problematic and 30pc were somewhat problematic. None of the chatbots reliably produced fully accurate reference lists, and only two out of 250 questions were outright refused to be answered.

Overall, the five chatbots performed roughly the same. Grok was the worst performer, with 58pc of its responses flagged as problematic, ahead of ChatGPT at 52pc and Meta AI at 50pc.

Performance varied by topic, though. Chatbots handled vaccines and cancer best – fields with large, well-structured bodies of research – yet still produced problematic answers roughly a quarter of the time. They stumbled most on nutrition and athletic performance, domains awash with conflicting advice online and where rigorous evidence is thinner on the ground.

Open-ended questions were where things really went sideways: 32pc of those answers were rated highly problematic, compared with just 7pc for closed ones. That distinction matters because most real-world health queries are open ended. People do not ask chatbots neat true-or-false questions. They ask things like: “Which supplements are best for overall health?” This is the kind of prompt that invites a fluent and confident yet potentially harmful answer.

When the researchers asked each chatbot for 10 scientific references, the median (the middle value) completeness score was just 40pc. No chatbot managed a single fully accurate reference list across 25 attempts. Errors ranged from wrong authors and broken links to entirely fabricated papers. This is a particular hazard because references look like proof. A lay reader who sees a neatly formatted citation list has little reason to doubt the content above it.

Why chatbots get things wrong

There’s a simple reason why chatbots get medical answers wrong. Language models do not know things. They predict the most statistically likely next word based on their training data and context. They do not weigh evidence or make value judgements. Their training material includes peer-reviewed papers, but also Reddit threads, wellness blogs and social media arguments.

The researchers did not ask neutral questions. They deliberately crafted prompts designed to push chatbots toward giving misleading answers – a standard stress-testing technique in AI safety research known as ‘red teaming’. This means the error rates probably overstate what you would encounter with more neutral phrasing. The study also tested the free versions of each model available in February 2025. Paid tiers and newer releases may perform better.

Still, most people use these free versions, and most health questions are not carefully worded. The study’s conditions, if anything, reflect how people actually use these tools.

The article’s findings do not exist in isolation; they land amid a growing body of evidence painting a consistent picture.

A February 2026 study in Nature Medicine showed something surprising. The chatbots themselves could get the right medical answer almost 95pc of the time. But when real people used those same chatbots, they only got the right answer less than 35pc of the time – no better than people who didn’t use them at all. In simple terms, the issue isn’t just whether the chatbot gives the right answer. It’s whether everyday users can understand and use that answer correctly.

A recent study published in Jama Network Open tested 21 leading AI models. The researchers asked them to work out possible medical diagnoses. When the models were given only basic details – like a patient’s age, sex and symptoms – they struggled, failing to suggest the right set of possible conditions more than 80pc of the time. Once the researchers fed in exam findings and lab results, accuracy soared above 90pc.

Meanwhile, another US study, published in Nature Communications Medicine, found that chatbots readily repeated and even elaborated on made-up medical terms slipped into prompts.

Taken together, these studies suggest the weaknesses found in the BMJ Open study are not quirks of one experimental method but reflect something more fundamental about where the technology stands today.

These chatbots are not going away, nor should they. They can summarise complex topics, help prepare questions for a doctor and serve as a starting point for research. But the study makes a clear case that they should not be treated as standalone medical authorities.

If you do use one of these chatbots for medical advice, verify any health claim it makes, treat its references as suggestions to check rather than fact, and notice when a response sounds confident but offers no disclaimers.

The Conversation

Carsten Eickhoff

Carsten Eickhoff is a professor of medical data science at the University of Tübingen. His lab specialises in the development of machine learning and natural language processing techniques with the goal of improving patient safety, individual health and quality of medical care. Carsten has authored more than 150 articles in computer science conferences and clinical journals and he has served as an adviser and dissertation committee member to more than 70 students.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.