ChatGPT wants your medical records – these are the risks

Dr ChatGPT will see you now. The artificial intelligence (AI) chatbot is launching a new health mode to give medical advice.

ChatGPT Health, which is not yet available in the UK, is a tab within the app where users can ask health-related questions.

The function was announced yesterday by OpenAI, the tech start-up behind the virtual assistant.

Health can analyse lab results, explain unclear messages from your doctor and provide your clinical history, OpenAI said.

Other usage examples include ‘post-surgery food guides’, health insurance provider picks or outlining drug side effects.

But the company stressed it’s not putting the ‘GP’ in ChatGPT, saying: ‘Health is designed to support, not replace, medical care. It is not intended for diagnosis or treatment.’

The company is encouraging users to connect their medical records and wellness apps, such as Apple Health, to get ‘personalised’ responses.

Patient paperwork will be encrypted, OpenAI added, and conversations won’t be used to train the chatbot.

When linking an app with the bot, ChatGPT says you’ll be in full control of how much access it has.

OpenAI said: ‘The first time you connect an app, we’ll help you understand what types of data may be collected by the third party. And you’re always in control: disconnect an app at any time and it immediately loses access.’

OpenAI said it worked with more than 260 physicians to provide feedback on health model outputs over the last two years, including teaching it to recommend follow-ups with a doctor when needed and not oversimplify.

Can you join the ChatGPT health waitlist?

ChatGPT Health is not available in the UK yet, with only a small group of early users able to test it.

Only people in the US can sign up for the waitlist right now.

The ability to upload medical records will only be available in the US – the UK and the European Union have strict privacy laws.

Using ChatGPT for health ‘could be dangerous’, says expert

Around one in four patients turn to AI to better understand their health, according to a survey by the digital health system Semble.

But medical experts have long expressed unease about this, warning it’s the latest version of people Googling health symptoms.

As much as the system can easily pass a medical licensing exam, researchers say it can spit out inaccurate or false information.

One man was taken to hospital with an 18th-century condition after taking ChatGPT’s alleged advice to replace salt in his diet with sodium bromide.

Chatbots also often try to please us by reinforcing what we say, meaning leading questions such as ‘Don’t you think I have the flu?’ could prompt it to agree with you, regardless of the facts, experts warn.

People, especially young adults, are also using AI systems as a form of therapy, to the alarm of psychological experts Metro spoke with.

Sophie McGarry, a solicitor at the medical negligence law firm Patient Claim Line, warned that AI health advice can be ‘very dangerous’.

She told Metro that bots may overdiagnose, underdiagnose or misdiagnose people.

‘This, in turn, could lead to potentially unnecessary stress and worry and could lead people to urgently seek medical attention from their GP, urgent care centres or A&E departments, which are already stretched, adding more unnecessary pressure or could lead to people attempting to treat their AI-diagnosed conditions themselves,’ she said.

‘As a clinical negligence solicitor, I see far too many cases of people’s lives being turned upside down because of misdiagnosis or delays in diagnosis where earlier, appropriate input would have led to a better, often life-changing, sometimes life-saving outcome,’ she said.

‘False reassurances from AI health advice could lead to the same devastating outcomes.’

Dray Agha, senior manager of security operations at Huntress, told Metro that the option in the US to upload private medical records is worrying.

Agha said: ‘While AI health tools can offer valuable insights, users probably shouldn’t upload their full medical records into any new platform, as it creates a significant and attractive target.

‘This concentrates sensitive, lifelong data in one place, and its security is only as strong as the platform’s safeguards against breaches and unauthorised access.’

Get in touch with our news team by emailing us at webnews@metro.co.uk.

For more stories like this, check our news page.

MORE: Dad and son both went to bed with ‘cough and cold’ before dying in their sleep

MORE: Elon Musk’s X could be banned in UK after Grok undressed women

MORE: NASA launches urgent evacuation of the ISS after ‘serious medical emergency’ in space