An artificial intelligence model trained on the medical data of 57 million people who have used the National Health Service in England could one day assist doctors in predicting disease or forecast hospitalisation rates, its creators have claimed. However, other researchers say there are still significant privacy and data protection concerns around such large-scale use of health data, while even the AI鈥檚 architects say they can鈥檛 guarantee that it won鈥檛 inadvertently reveal sensitive patient data.

The model, called Foresight, was first developed in 2023. That initial version used OpenAI鈥檚 GPT-3, the large language model (LLM) behind the first version of ChatGPT, and trained on 1.5 million real patient records from two London hospitals.

Now, at University College London and his colleagues have scaled up Foresight to create what they say is the world鈥檚 first 鈥渘ational-scale generative AI model of health data鈥� and the largest of its kind.

Foresight uses eight different datasets of medical information routinely collected by the NHS in England between November 2018 to December 2023 and is based on Meta鈥檚 open-source LLM Llama 2. These datasets include outpatient appointments, hospital visits, vaccination data and records, comprising a total of 10 billion different health events for 57 million people 鈥� essentially everyone in England.

Tomlinson says his team isn鈥檛 releasing information about how well Foresight performs because the model is still being tested, but he claims it could one day be used to do everything from making individual diagnoses to predicting broad future health trends, such as hospitalisations or heart attacks. 鈥淭he real potential of Foresight is to predict disease complications before they happen, giving us a valuable window to intervene early, and enabling a shift towards more preventative healthcare at scale,鈥� he told a press conference on 6 May.

While the potential benefits are yet to be supported, there are already concerns about people鈥檚 medical data being fed to an AI at such a large scale. The researchers insist all records were 鈥渄e-identified鈥� before being used to train the AI, but the risks of someone being able to use patterns in the data to re-identify the records are well-recorded, particularly when it comes to large datasets.

Free newsletter

Sign up to The Daily

Sign up to newsletter

鈥淏uilding powerful generative AI models that protect patient privacy is an open, unsolved scientific problem,鈥� says at the University of Oxford. 鈥淭he very richness of data that makes it valuable for AI also makes it incredibly hard to anonymise. These models should remain under strict NHS control where they can be safely used.鈥�

鈥淭he data that goes into the model is de-identified, so the direct identifiers are removed,鈥� said at NHS Digital, speaking at the press conference. But Chapman, who oversees the data used to train Foresight, admitted that there is always a risk of re-identification: 鈥淚t鈥檚 then very hard with rich health data to give 100 per cent certainty that somebody couldn鈥檛 be spotted in that dataset.鈥�

To mitigate this risk, Chapman said the AI is operating within a custom-built 鈥渟ecure鈥� NHS data environment to ensure that information isn鈥檛 leaked out of the model and is accessible only to approved researchers. Amazon Web Services and data company Databricks have also supplied 鈥渃omputational infrastructure鈥�, but can鈥檛 access the data, said Tomlinson.

at Imperial College London says one way to check whether models can reveal sensitive information is to verify whether they can memorise data seen during training. When asked by New 女生小视频 whether the Foresight team had conducted these tests, Tomlinson said it hadn鈥檛, but that it was looking at doing so in the future.

Using such a vast dataset without communicating to people how the data has been used can also weaken public trust, says at the University of Oxford. 鈥淓ven if it is being anonymised, it鈥檚 something that people feel very strongly about from an ethical point of view, because people usually want to keep control over their data and they want to know where it鈥檚 going.鈥�

But existing controls give people little chance to opt out of their data being used by Foresight. All of the data used to train the model comes from nationally collected NHS datasets, and because it has been 鈥渄e-identified鈥�, , says an NHS England spokesperson, though people who have chosen not to share data from their family doctor won鈥檛 have this fed into the model.

Under the General Data Protection Regulation (GDPR), people must have the option to withdraw consent for the use of their personal data, but because of the way LLMs like Foresight are trained, it isn鈥檛 possible to remove a single record from an AI tool. The NHS England spokesperson says that 鈥渁s the data used to train the model is anonymised, it is not using personal data and GDPR would therefore not apply鈥�.

Exactly how the GDPR should address the impossibility of removing data from an LLM is an , but the UK Information Commissioner鈥檚 Office鈥檚 website states that 鈥渄e-identified鈥� data should not be used as a synonym for anonymous data. 鈥淭his is because UK data protection law doesn鈥檛 define the term, so using it can lead to confusion,鈥� it states.

The legal position is further complicated because Foresight is currently being used only for research related to covid-19, says Tomlinson. That means exceptions to data protection laws enacted during the pandemic still apply, says Sam Smith at , a UK data privacy organisation. 鈥淭his covid-only AI almost certainly has patient data embedded in it, which cannot be let out of the lab,鈥� he says. 鈥淧atients should have control over how their data is used.鈥�

Ultimately, the competing rights and responsibilities around using medical data for AI leave Foresight in an uncertain position. 鈥淭here is a bit of a problem when it comes to AI development, where the ethics and people are a second thought, rather than the starting point,鈥� says Green. 鈥淏ut what we need is the humans and the ethics need to be the starting point, and then comes the technology.鈥�

A US law banning the popular video-sharing app TikTok is set to take effect at the start of 2025 鈥� but the US Supreme Court has to hear TikTok鈥檚 legal challenge to this. Meanwhile, President-elect Donald Trump has hinted he might take action against the law, raising new questions about whether it will stand.

What would the TikTok ban actually do?

Starting on 19 January 2025, the 鈥溾�� would block US companies such as Google and Apple from allowing users to access or update TikTok through their app stores 鈥� unless TikTok鈥檚 Chinese owner ByteDance sells the app to a US company. It would also require internet service providers to block the platform on US internet browsers. It was approved with bipartisan support by the House of Representatives and Senate before being signed into law by President Joe Biden in April 2024.

If the ban is implemented, it would be practically impossible for new users in the US to download the TikTok app, says at the Center for Democracy & Technology, a non-profit organisation based in Washington DC. For the 170 million existing TikTok users in the US, the app might remain on their phones. Without access to updates, however, its functionality would degrade over time.

People in the US could still potentially access TikTok using a virtual private network (VPN) service that disguises a user鈥檚 location. But the experience of using the app could still worsen, says Ruane: because TikTok content would no longer be housed on nearby US servers, it would load more slowly.

These restrictions stem from concerns about privacy and security. US lawmakers have stated that TikTok is a 鈥渘ational security threat鈥� because China鈥檚 government could force ByteDance to hand over TikTok users鈥� data or pressure the app to modify its algorithm, presenting content that could manipulate public opinion. However, no firm evidence has been provided to support these claims. TikTok has said 聽from outside influence and manipulation.

鈥淚t鈥檚 very concerning that a country like the United States, which has consistently led on the global stage in defending the free, open and interoperable internet, is now taking a step to ban access to an entire platform within its borders 鈥� that is an extraordinary measure,鈥� says Ruane.

Free newsletter

Sign up to The Weekly

Sign up to newsletter

Will the Supreme Court block the TikTok ban?

Although judges in the lower DC Circuit Court of Appeals previously the US law to stand, the Supreme Court has agreed to hear TikTok鈥檚 appeal. TikTok鈥檚 is that the ban amounts to censorship that violates the free speech rights afforded to Americans under the US Constitution鈥檚 First Amendment.

鈥淚 would like to see the court really grapple with how this law harms those rights, and how the government should account for the rights of social media users when it is attempting to regulate these speech platforms,鈥� says Ruane. 鈥淭he courts have not done that in the course of examining this particular law, even though there are users who are suing, claiming that the law violates their First Amendment rights, as distinct from TikTok.鈥�

The most likely short-term impact is that the US Supreme Court will temporarily suspend implementation of the law while the justices are considering the case, says Ruane. That could delay the law鈥檚 effects for months 鈥揾owever long the Supreme Court requires to make its decision in 2025. TikTok has specifically asked for such a hiatus in its court filing.

If the Supreme Court finds that the ban infringes First Amendment rights, and that the US government has less restrictive options at its disposal, it could issue an injunction that effectively makes it impossible for the government to justify such an outright ban, says Ruane. The Supreme Court may also require the lower DC Circuit Court of Appeals to reexamine its analysis of the case. Such decisions could force the government to find more narrowly tailored options for regulating TikTok.

How could Trump prevent the TikTok ban?

In his first term, President-elect Trump supported plans to ban TikTok, but he has since changed his stance. During the 2024 presidential campaign, he promised to 鈥溾�� in a post on his social media platform, Truth Social, as he encouraged US voters to back him. On 16 December, Trump the CEO of TikTok and later, during a press conference, said his administration would 鈥渢ake a look鈥� at the ban. Even if the Supreme Court eventually agrees to let the ban stand, Trump could change the impact of the law.

For example, the president could go back to US lawmakers and ask them to change the national law by repealing or amending it, says Ruane. She also described a scenario where Trump could potentially direct his administration鈥檚 US attorney general to not enforce the law 鈥� although she cautioned that would be outside the norms of how the US government has typically operated.

Even if Trump鈥檚 attorney general announced that the US government wouldn鈥檛 enforce the ban, US companies such as Google and Apple might still be reluctant to allow people to access the app through their platforms. 鈥淚f I am in charge of legal risk at one of these companies, I don鈥檛 know whether I would say 鈥榳e鈥檙e going to trust that [decision], it鈥檚 fine for us to allow access to this app that has been banned,鈥欌�� says Ruane.

What would a US TikTok ban mean for the rest of the world?

If it goes through, the US ban could have significant ripple effects across the world. For starters, people in other countries would no longer have access to new content from US-based TikTok creators and influencers. But more crucially, the US government鈥檚 actions could encourage other countries eying similar restrictions.

The US isn鈥檛 the first nation to act against TikTok 鈥� India鈥檚 government has blocked the app since 2020 鈥� but Ruane expressed concerns that a US ban could encourage 鈥渁uthoritarian regimes鈥� to bar any app, including some developed in the US, by invoking similar national security justifications.

鈥淚 think it absolutely will be used as a justification for banning TikTok elsewhere, but also for banning access to other applications which have served as important speech platforms in countries where the internet might not otherwise be as open,鈥� says Ruane.

Would the TikTok ban protect privacy?

The ostensible goal of the ban is to protect the privacy of US TikTok users 鈥� preventing their data from falling into the hands of another nation 鈥� and to address concerns that China鈥檚 government could manipulate content presented to app users in the US. But Ruane says there are many alternative actions US lawmakers could take before blocking TikTok entirely.

For instance, the government could require TikTok to be more transparent about how it gathers and shares individual users鈥� data, and what measures it takes to protect their privacy. To alleviate concerns of manipulation, lawmakers might require the platform to share how its algorithms filter and manage the content that users see, says Ruane.

The US government could also consider passing a consumer privacy law to provide better legal protections for how social media platforms can share individuals鈥� data with other companies or governments. 鈥淭hose options regarding consumer privacy and transparency are less extreme than banning an entire platform,鈥� says Ruane.