Slingshot AI, a well-funded startup behind the therapy-focused chatbot Ash, withdrew the app from the United Kingdom in January 2026 after concluding there was no clear regulatory path for a product straddling the line between a wellness tool and a medical device. Five months later, the company shifted strategy again, launching a clinical research initiative that pairs its chatbot with real, licensed human therapists — a reversal that illustrates just how unsettled the rules still are for AI products that market themselves as therapeutic.
A Fast, Well-Funded Launch
Ash launched in July 2025 as a free iOS and Android app after eighteen months of development and testing with roughly 50,000 beta users, backed by $93 million in funding from investors including Andreessen Horowitz. The company’s pitch was ambitious: Slingshot described Ash as the first AI model built specifically for therapy, aiming to give users "deeply personalized mental health support" from a chatbot available on any phone, at any hour, often at far lower cost than seeing a human therapist. The company positioned itself at the front of a crowded field of AI mental-health apps — alongside established consumer tools like Woebot, Wysa, and Replika — but marketed Ash more explicitly as a therapeutic product rather than a general wellness or companionship app, a framing that would soon run into trouble overseas.
The UK Pulls the Plug
According to a STAT News report dated January 21, 2026, Slingshot told UK users by email that it would withdraw Ash from the country that week, with the app becoming unavailable after January 23. The core issue was regulatory classification: the UK’s Medicines and Healthcare products Regulatory Agency treats software that performs functions associated with diagnosing or treating medical conditions, including certain mental health conditions, as a regulated medical device requiring clearance, not a consumer wellness app that can launch freely. Slingshot’s position was that Ash is a "wellbeing" product, not a medical device, but the company apparently could not get clarity or approval under that framing. Slingshot CEO Daniel Reid Cahn wrote to users that "there isn’t a clear regulatory pathway for wellbeing products like ours," adding that without that clarity, the company couldn’t operate with confidence. Slingshot said it remained in conversations with the UK government about a possible remedy, though STAT reported the company did not respond to further requests for comment at the time.
Part of a Broader Pattern
STAT’s reporting framed Slingshot as one of the most aggressive companies pushing large-language-model-based mental health products directly to consumers, at a moment of rising, broader concern that generative AI chatbots can pose real dangers for people experiencing mental health crises. That concern has shown up elsewhere in 2026: Illinois, Nevada, and Utah have all passed laws restricting how AI can be marketed or used for mental health services domestically, with Illinois barring AI from making treatment decisions or communicating directly with clients in a therapeutic capacity, and Nevada imposing civil penalties of up to $15,000 for AI systems that present themselves as capable of professional behavioral health care. Slingshot’s UK exit suggests the regulatory friction isn’t confined to the United States, and that European and UK regulators may apply medical-device frameworks more aggressively to chatbots that market themselves using clinical, therapy-adjacent language.
The Pivot: Put a Human in the Loop
By May 2026, Slingshot had changed its public strategy. A company job posting from that month describes a new clinical research initiative designed to study how Ash, used alongside real therapy, affects patient outcomes — seeking licensed clinicians for defined, four-week patient engagements with a specific clinical scope, based out of offices in New York City and London. The move effectively walks back the earlier framing of Ash as a self-contained AI substitute and instead repositions the product as a complement to licensed human care, generating the kind of outcomes data that regulators and skeptical clinicians have been asking for. It’s a strategy that mirrors what Dartmouth researchers did with their own AI therapy chatbot, Therabot, which was studied in a human-monitored randomized controlled trial rather than released for unsupervised public use.
Skeptics Remain Unconvinced
Mental health professionals have been broadly cautious about chatbots marketed as therapy substitutes, even well-funded ones. A 2026 American Psychological Association survey of 1,242 licensed psychologists found that 94% did not believe current AI chatbots could treat psychological conditions with adequate nuance, and 97% worried chatbots could reinforce negative behaviors or delusional thinking rather than challenge them — concerns that predate Ash specifically but apply directly to any product using therapeutic language without clinical supervision. Critics argue that a company reversing from unsupervised consumer deployment to clinician-paired pilots, after a regulator effectively forced the issue, raises the question of why that safeguard wasn’t built in from the start.
What’s Next
Slingshot’s experience is likely to be studied closely by competitors and regulators alike as a test case for how AI therapy products can — or can’t — scale globally. If the UK and other regulators settle on a clear wellness-versus-medical-device line, it could either open a path for Ash’s return or push the company and its rivals to seek formal medical-device clearance, a slower and more expensive route but one that would give the product more regulatory durability. In the meantime, Slingshot’s shift toward clinician-paired trials signals a broader industry recognition that, at least for now, the fastest way to prove an AI therapy product is both safe and effective may be to put a licensed human back in the loop rather than remove one.
Photo: Shantanu Kumar / PEXELS via Pexels
