Hi friend,

If you’re new here, welcome! Every week I write this email to keep you up to date on what’s happening in mental health innovation. I collect the latest funding rounds, industry and policy news, frontier research and more. Then I read them, filter them, summarise them and include links in case you want to read more. I hope you find it valuable.

And if you’ve been around for a while, well, welcome back! It’s great to have you here!

Now let’s get into it!

⏺️ Jimini Health raised $17M. Jiminy develop patient facing AI solutions that are delivered through hybrid AI-human care models. (link)

⏺️ Headway acquired Tezi, an AI workflow automation company. This was an acqui-hire (an acquisition for the talent of the team) which Alma will use to accelerate its AI capabilities across billing, insurance, and care navigation. (link)

⏺️ Neurolief secured $6M. The milestone-stone based investment from Brainsway, followed the FDA approval of Proliv Rx - an at-home brain neuromodulation device for TRD - earlier this year. The funding will support Neurolief’s US rollout. (link)

⏺️ The Ballmer Group donated $110M across three Los Angeles universities to expand youth mental health services. Cal State LA receives $48M, UCLA and Cal State Dominguez Hills share the remainder. (link)

⏺️ NorthBay Health received $14.9M through California's Behavioural Health Continuum Infrastructure Program to develop a 24/7 behavioural health and addiction centre. (link)

⏺️ In the US, the VA selected Mynd Immersive to lead a national rollout of prescriptive virtual reality therapies for veterans. The programme covers chronic pain, anxiety, stress, and isolation. (link)

⏺️ SimplePractice and Uprise Health announced a partnership to expand mental health access via SimpleConnect. The integration connects Uprise Health's EAP members directly to independent practitioners on SimplePractice's platform via a matching engine, enabling direct-to-calendar booking. (link)

⏺️ Alma laid off 45 employees as it works through its merger with Spring Health, The cuts affected entry-level operations and customer experience roles. Thier pending acquisition by Spring is expected to close in Q2 2026. (link)

⏺️ US Federal agencies announced plans to revise Mental Health Parity and Addiction Equity Act requirements. The HHS, Treasury, and Labor Department are developing new rules that could affect how insurers and employers structure behavioural health benefits. (link)

⏺️ Utah launched a one-year pilot allowing an AI chatbot from Legion Health to autonomously renew 15 low-risk psychiatric prescriptions. The pilot requires physician review of the first 1,250 requests and excludes controlled substances, antipsychotics, and dose changes. (link)

⏺️ Tennessee enacted a law prohibiting AI systems from representing themselves as qualified mental health professionals. The legislation includes a private right of action for violations. (link)

⏺️ The UK government published an interim report from its independent review into mental health conditions, ADHD, and autism. The review found common mental health disorders have increased - particularly among younger adults - while rising diagnosis rates for ADHD and autism reflect improved recognition rather than rising prevalence. Final findings are due summer 2026. This is a super interesting report that I would recommend reading. (link)

If you find these updates valuable and would like to connect with other leaders in mental health innovation, consider joining the Hemingway Community.

As a member of the community, you’ll join over 420 other mental health leaders and get access to exclusive content, events and resources. We are hosting in-person events in SF, Chicago Sydney and DC in the next few months as well as expert-led online events with leaders like Dana Udall PhD and more. You’ll also get invited to our vetted Slack group.

⏺️ When Ash, a mental health AI app, was withdrawn from the UK market, 53% of former users turned to general-purpose AI or stopped seeking support entirely. The preprint from Slingshot found fewer than one in eight users transitioned to professional care, and 30% sought no support at all. In this group, 81% said their access had worsened and 49% reported worse-than-usual wellbeing in the past month. (link)

⏺️ A Science study found that sycophantic AI reduces prosocial behaviour and increases user dependence. Across 11 LLMs and preregistered experiments with 1,604 participants, even a single interaction with a sycophantic AI increased users' conviction of being correct and reduced willingness to repair interpersonal harm. Despite distorting judgment, sycophantic models were trusted and preferred (link)

⏺️ Verily Life Sciences published a study in npj Digital Medicine testing an AI guardrail system for detecting psychiatric crises in text-based conversations. Tested against two clinician-labelled datasets totalling nearly 2,600 messages, the system achieved sensitivity of 0.990 and specificity of 0.992 on Verily's own dataset, and outperformed guardrails from NVIDIA and OpenAI on sensitivity. The majority of authors are Verily employees evaluating their own tool, which should be weighed when interpreting the results. (link)

⏺️ Flourish Science published a company blog post reporting that its mental health AI Sunnie scored 86/100 on VERA-MH, the open source AI safety benchmark from Spring. The post goes into good details on how Flourish approach safety and is an interesting read for anyone interested in this field. (link)

⏺️ A nationally representative survey of 2,383 young people aged 13–24 found that feeling genuinely seen in human relationships was one of the strongest protections against high-risk AI use. The Rithm Project report identified four clusters of AI engagement among young people and found that relational quality — not just individual traits — shaped how they used and were affected by AI. (link)

⏺️ A randomised trial found that TMS targeting a novel brain circuit outperformed the standard target for anxiety symptoms in patients with comorbid anxiety and depression. The Molecular Psychiatry trial (n=40) found anxiety improvement of 58% vs 36% for the standard target, while both targets performed equally for depression. (link)

⏺️ A qualitative study of 18 psychotherapists found that trust in generative AI depends on whether it preserves professional role integrity and human connection. Clinicians supported AI for documentation and brainstorming but resisted tools perceived as threatening the therapeutic relationship. (link)

⏺️ A meta-analysis of 24 studies found machine learning can predict CBT outcomes for depression with moderate accuracy (r=0.45). Depression severity, functional impairment, and sociodemographic factors were the strongest predictors. (link)

⏺️ A scoping review found emerging evidence that GLP-1 receptor agonists may benefit substance use disorders and eating disorders. The strongest data comes from real-world pharmacovigilance studies; the authors call for randomised trials to confirm psychiatric applications. (link)

⏺️ A large-scale genome-wide association study identified 25 novel genetic loci linked to anorexia nervosa, including one near the SOX5 gene. The findings add to evidence for a biological basis of eating disorders and point to complex psychiatric-metabolic genetic pathways. (link)

⏺️ Analysis of a three-arm OCD trial found that reductions in "inferential confusion" — treating imagined possibilities as reality — predicted symptom improvement across CBT, inference-based CBT, and mindfulness approaches. The finding points to a shared mechanism underlying different OCD treatments. (link)

That’s all for this week.

Keep fighting the good fight!

Steve Duke

Founder of Hemingway

P.S. Find this valuable and want to support our work? Consider becoming a Hemingway Pro member.