This month, two of the most popular AI firms in San Francisco introduced a serious push into healthcare — strikes that specialists say weren’t solely inevitable, but in addition well timed and high-stakes.
These AI rivals — Anthropic and OpenAIthe makers of the broadly used massive language fashions Claude and ChatGPT, respectively — unveiled new suites of instruments for healthcare organizations and on a regular basis shoppers. These strikes mirror a shift in how sufferers are accessing medical steerage — one which specialists agree is concurrently increasing entry to data whereas elevating new questions on belief and management.
What these healthcare expansions might imply for startups
Anthropic and OpenAI’s healthcare buildouts are forcing startups throughout the well being tech market to reassess the place they really have defensible benefits, one investor identified.
Kamal Singh, senior vp at WestBridge Capitalthinks shopper wellness and diet startups are probably the most susceptible, saying that these kind of broad, chat-based platforms are prone to be commoditized.
Startups providing diet or wellness recommendation with out deep specialization now face weakened worth propositions — provided that Claude and ChatGPT have huge distribution and routine utilization, he identified. Some examples embody apps like Noom, Fay and Zoe.
Others will in all probability stay insulated — and even strengthened — relying on how strong their fashions are, Singh stated. In his view, firms centered on specialised scientific areas, reminiscent of continual illness administration, can be much more resilient to massive tech incumbents getting into the area.
Some of these firms depend on deep affected person knowledge, longitudinal insights and disease-specific experience — capabilities that we nonetheless don’t know if basic function tech firms will have the ability to replicate at scale, Singh remarked.
He additionally pointed to care coordination and care administration as areas the place startups can keep an edge, significantly once they mix AI with human clinicians. Reasonably than competing immediately with massive language fashions, Singh believes startups ought to differentiate by prioritizing outcomes and delivering end-to-end care experiences.
One other rising battleground is AI-driven major care. Singh stated this class sits between shopper wellness and specialised medication — subtle sufficient to withstand full commoditization, however nonetheless susceptible to strain from widespread AI platforms.
“On the startup facet, you don’t actually have any winners but — there are a few firms like Counsel Well being, who’re form of inching in the direction of that purpose, however these bulletins make it a really fascinating dynamic there,” he declared.
Counsel Well being is a digital care firm that mixes AI with human physicians to offer customers fast, customized medical recommendation.
To outlive, Singh stated startups on this area will want inventive enterprise fashions, together with hybrid approaches that combine actual clinicians with AI-powered steerage.
The inevitable rise of AI as healthcare’s entrance door
It was inevitable that OpenAI and Anthropic would deepen their presence in healthcare. Traits in person exercise made this unavoidable — lots of of thousands and thousands of individuals per week had been turning to their chatbots to reply their health-related inquiries.
“Nearly 5% of their site visitors is healthcare-related. There are about 40 million distinctive healthcare questions requested by customers in a day. On condition that, it actually does appear that they’re within the healthcare enterprise, and so in the event that they’re seeing that a lot site visitors to their websites associated to healthcare, they needed to improve their capabilities in that area,” defined healthcare AI skilled Saurabh Gombar.
So what did the Anthropic and OpenAI truly roll out?
OpenA launched two new choices. One is ChatGPT Well beinga devoted well being expertise inside ChatGPT that mixes a person’s private well being data with the corporate’s AI, with the promise of serving to folks higher handle their well being and wellness. The opposite is OpenAI for Healthcarea set of AI instruments designed to assist healthcare suppliers cut back administrative burnout and enhance care planning.
OpenAI additionally introduced its acquisition of medical data startup Torch this month — a deal that’s reportedly value $100 million.
Anthropic adopted with a healthcare splash of its personal, unveiling a new suite of Claude instruments. The corporate is releasing new agent capabilities for duties like prior authorization, healthcare billing and scientific trial workflows, in addition to letting its paid customers join and question their private medical data to get summaries, explanations and steerage for physician visits.
Gombar, the AI skilled talked about above, believes that giant language fashions have gotten the brand new “entrance door” to healthcare.
“The LLMS are actually turning into the entrance door for medical recommendation and remedy choices, and the precise supplier is turning into the second opinion. As a result of chatbots are simpler to work together with, and so they’re free, and also you don’t must schedule round them,” Gombar acknowledged.
Gombar is a scientific teacher at Stanford Well being Care and chief medical officer and co-founder of Atropos Well beinga healthcare AI startup that generates real-world proof on the bedside. In his eyes, tech firms creating public-facing chatbots are already within the healthcare enterprise, whether or not they formally acknowledge it or not.
This might basically alter the physician-patient relationship. Gombar famous that clinicians are already starting to see an increasing number of sufferers who arrive already satisfied they want particular checks or remedies based mostly on chatbot recommendation.
He thinks conventional suppliers have restricted management over this shift, given shopper conduct is clearly altering at a speedy tempo. Not solely has using chatbots like ChatGPT and Claude skyrocketed previously couple of years, however People are additionally discovering it harder to entry healthcare amid sweeping Medicaid cuts and a worsening labor scarcity.
The dangers of chatbots in medication
The rise of huge language fashions in healthcare is already properly underway, however that doesn’t imply there aren’t dangers concerned. Asking for medical steerage from an clever software program program could be very completely different than asking for a recipe — unsuitable solutions can trigger actual hurt.
Conventional healthcare suppliers have accountability mechanisms — reminiscent of medical malpractice guidelines, audit trails and legal responsibility protocols — whereas chatbots rely closely on disclaimers that say their outputs shouldn’t be thought-about medical recommendation, Gombar identified.
Nevertheless, in apply, many customers deal with chatbot responses as precise medical recommendation, usually with out cross-checking with different sources or their suppliers, he added.
Gombar hopes firms like Anthropic and OpenAI transfer past disclaimers and take larger duty for the way their instruments deal with medical data. Sooner or later, he want to see them be extra clear concerning the limitations of their programs — together with how usually they hallucinate, when solutions will not be grounded in robust proof and when medical proof itself is unsure or incomplete.
He additionally recommended that giant language fashions be designed to extra clearly talk uncertainty and gaps in data, quite than presenting speculative solutions with unwarranted confidence, he stated.
Except for accuracy, there are additionally issues associated to knowledge privateness, as shoppers’ rising mistrust of Massive Tech firms and their knowledge privateness practices stays an ongoing difficulty.
Anthropic stated that its new well being merchandise are designed with strict safeguards round person consent and knowledge safety.
“Customers give categorical consent to combine their knowledge with full details about how Anthropic protects that knowledge in our shopper well being knowledge privateness coverage. Anthropic doesn’t practice on person well being knowledge. Interval. We additionally defend delicate well being knowledge from inadvertent sharing to different built-in mannequin context protocols by requiring person consent to every integration in conversations the place built-in well being knowledge is being mentioned. Customers can disconnect the mixing any time in settings,” an Anthropic spokesperson defined in an emailed assertion.
Even earlier than it rolled out ChatGPT Well being, OpenAI had been constructing person knowledge protections throughout ChatGPT, together with everlasting deletion of chats from OpenAI’s programs inside 30 days and coaching its fashions to not retain private data from person chats, an organization spokesperson stated in a press release.
For its new shopper well being providing, OpenAI has added extra encryption protections, in addition to remoted the chats to maintain well being conversations and reminiscence protected and compartmentalized. Conversations in ChatGPT Well being will not be used to coach its basis fashions, the spokesperson stated.
As for OpenAI’s new platform for healthcare suppliers, prospects could have full management over their knowledge. When clinicians enter affected person data, for instance, it would keep throughout the group’s safe workspace and won’t be used for mannequin coaching.
Making AI work for clinicians and sufferers
By releasing instruments for shoppers in addition to for healthcare suppliers, OpenAI is signaling that it understands shoppers have completely different wants and objectives than hospitals. Sufferers need basic steerage and comfort, whereas suppliers want correct, actionable data that may be safely built-in into the scientific document, famous Kevin Erdal, senior vp of transformation and innovation providers at Nordica well being and know-how consultancy.
When deploying new massive language fashions, he really useful hospitals be careful for shadow workflows.
“Clinicians might begin informally counting on patient-generated summaries or AI-assisted interpretations with out clear requirements for validation or documentation. If nobody validates the place patient-reported data got here from, or oversees how that data is reviewed, integrated or rejected, threat quietly accumulates,” Erdal stated.
In terms of Anthropic and OpenAI’s consumer-facing healthcare instruments, the most important threat isn’t misinformation a lot as lacking context, he remarked.
“Context, intent and reasoning can stay in a chat whereas the scientific document captures solely the end result, weakening care continuity and the belief between affected person and supplier,” Erdal acknowledged.
This hole in context underscores why consumer-facing chatbots are ill-suited for clinician use.
For hospitals and different suppliers, Erdal thinks the fitting response to the rise of consumer-facing healthcare AI is integration.
“It would appear to be well being programs accepting that these instruments exist already, and designing accountable methods to soak up their output with out fragmenting care. The bar is continuity, and the affected person/supplier relationship is what’s at stake,” he declared.
If consumer-facing AI fashions assist sufferers stroll into healthcare interactions extra knowledgeable and higher ready, however then their suppliers are unprepared to combine that into the healthcare dialog in a considerate or deliberate method, entry to healthcare data improves whereas belief drops off, Erdal defined.
At a deeper degree, OpenAI and Anthropic’s healthcare push displays a broader shift within the healthcare business.
The query is not whether or not AI will turn out to be a part of the affected person journey — it’s clear that the shift is already underway. The true query is who will management it, who can be accountable for it, and the way a lot affect it would have over choices that had been as soon as firmly within the arms of clinicians.
Consultants agree that the businesses that adapt — by integrating AI thoughtfully, strengthening belief and clarifying duty — might assist construct a extra accessible healthcare system. Those who don’t might discover themselves left behind.
Photograph: Pakorn Supajitsoontorn, Getty Photographs
