Hi friend,

There was some major mental health infrastructure news this week, ranging from research facilities in Oxford to a $1B hospital contract in Riyadh.

Pelago, Klarify and Alleva all launched new products, as did Anthropic, which released Claude Science.

On the research front, new work from John Torous and colleagues estimates that 27% of people use AI for their mental health, Slingshot share real-world outcomes from users of Ash, and birdsong is found to have positive effects on wellbeing.

Let’s get into it.

Come meet other Hemingway Members this Thursday. We’re hosting a Hemingway Session with Jas Schembri-Stothart, Co-Founder of Luna, and we’d love you to come along. Luna is the UK's leading health and wellbeing app for teen girls, and Jas will be running a session on how to build mental health products for young people (ones they actually use). Come along, learn from Jas and get a chance to connect with other members of the Hemingway Community.

RSVP here > https://luma.com/hemingway-bbui

⏺️ A Saudi-German consortium won a $1B mental health hospital contract. The Saudi government awarded the operation and management contract for the SABIC Specialised Behavioural Healthcare Hospital in North Riyadh to the SEH Healthcare joint venture. The total contract is valued at over $1B across fifteen years. The public-private partnership is one of the largest single mental health infrastructure investments ever in the MENA region, reflecting accelerating investment and privatisation of healthcare as part of Saudi Arabia’s Vision 2030. (link)

⏺️ BRYM raised €650K to build a wearable neurofeedback platform. The Stockholm-based company is developing a device that uses EEG to improve cognitive performance. (link)

⏺️ XTL BioPharma completed its acquisition of Psyga Bio and pivoted to psychedelic therapies. The deal shifts XTL’s strategy to focus on psychedelic medicine, the area in which Psyga Bio holds a clinical pipeline of candidates. XTL also completed a $1.5M raise alongside the transaction. (link)

⏺️ Oxford City Council green-lights a £750M mental health and research campus. Oxford Health NHS Foundation Trust and the University of Oxford received planning approval for Warneford Park, which will replace the 200-year-old Warneford Hospital with a new inpatient facility, a brain and mental health research centre, and biotech workspace. Construction could begin as early as 2028. (link)

⏺️ Pelago launched Sona, a voice-first AI mental health coach for employees. Sona will be available at no cost to eligible employees through their employer and is designed for everyday stress, burnout, and emotional challenges, with a referral pathway to clinical care when needed. (link)

⏺️ Klarify launched an AI operating system for therapists. The Y Combinator-backed company joins a long list of businesses attempting to reduce administrative burden and insurance complexity for private-practice clinicians. (link)

⏺️ Alleva launched Alleva Intelligence, an AI suite for behavioural health operators. While Klarify is focusing on private-practice clinicians, Alleva targets larger behavioural health organisations, attempting to reduce administrative burden and improve compliance, operational visibility and clinical insights. (link)

⏺️ Anthropic launched Claude Science, an AI workbench for pharmaceutical researchers. One of the big promises of AI was that it would help us do better science and make discoveries that help humanity. Let’s see if Claude Science delivers on that promise. (link)

⏺️ Connecticut classified brain data as sensitive personal information from July 1. EEG readings from consumer wearables are now covered under the state's privacy law, creating new consent and data handling obligations for companies building neuromodulation or mental health wearable products. (link)

⏺️ The US Congress introduced a bill to require insurers to cover three no-cost behavioural health visits. The Primary and Behavioral Health Care Access Act of 2026 would require private insurers to provide three outpatient mental health or substance use visits, with zero co-pay, each year. The bill has not yet advanced through the Senate. (link)

⏺️ Bipartisan children's online safety legislation passed the US House. The bill addresses platform regulation with direct implications for youth mental health and will now advance to the Senate. (link)

⏺️ Minnesota now requires social media platforms to show mental health warnings. Under an amendment to the state's consumer data privacy law, platforms must now display pop-up mental health warnings before users reach their feeds. Tech companies are fighting the requirement in federal court. (link)

⏺️ New Zealand passed a landmark mental health bill. The changes to the Mental Health Act govern how people can be assessed and treated. Commentators have referred to the bill as the biggest law reform for mental health in New Zealand in over 30 years. (link)

⏺️ A California man sued OpenAI, claiming ChatGPT worsened his mental health. The lawsuit challenges OpenAI's positioning of its AI as a health and wellness tool and represents an early test of AI liability in mental health contexts. (link)

⏺️ Review estimates around 27% of AI users have used AI for mental health support. How many people actually use AI for mental health? Torous and colleagues set out to find an answer. They reviewed 19 surveys published since 2024 and found a wide range of results, from 3%–70% across studies. The wide variation reflects the absence of agreed definitions for terms like therapy, emotional support, and wellness. Based on their findings, the authors approximate that 27% of AI users utilise AI for mental health support and highlight that around 40% of AI users said they were open to using AI for mental health support. Get ready to start seeing that 27% number everywhere. (link)

⏺️ Slingshot published real-world outcomes from Ash, their mental health AI tool. A single-arm observational study of 1,284 users found significant within-person improvements in life satisfaction, relationship satisfaction, sleep quality, and behavioural activation over four weeks (d = 0.14–0.26). More active days and sessions predicted better outcomes; message volume alone did not. The study had no control group and is not yet peer-reviewed. (link)

⏺️ New report from Imperial College found children under 13 are an underserved and high-risk group in UK mental health services. The New Childhood report, produced by Imperial's Institute of Global Health Innovation and Mental Health Innovations, found that children aged 13 and under account for up to 15% of all Shout text crisis conversations. Over 90% returned to the service more than once. Conversations about suicidality in this age group have risen since 2023 despite an overall decline across the broader user base. One in five children aged 8 to 16 has a probable mental health disorder, with over half of all mental health issues estimated to start before age 14. The report found that existing services are designed for older adolescents and calls for lower-threshold, developmentally attuned support earlier in childhood. (link)

⏺️ New research shows real-world outcomes for adolescents who engaged with Grow Therapy clinicians. A retrospective cohort study of nearly 10,000 adolescents aged 9 to 17 found that over half achieved a clinically meaningful improvement in depression (51.3%) and anxiety (57.8%) after engaging with a Grow therapist. The study also found that females were marginally more likely to attend three or more sessions but less likely to achieve a clinically meaningful improvement than males. Females also accounted for three in four of all adolescents engaged in care. (link)

⏺️ Large real-world study found Collaborative Care works for severe depression, not just moderate cases. Researchers analysed 30,162 closed patient episodes at Concert Health between 2018 and 2025, including 5,693 patients who started with severe depression (PHQ-9 above 20). Around 52% of patients with severe depression achieved a clinically meaningful response, a similar rate to patients with moderate depression. Remission was harder to achieve as severity increased, falling from 42% in moderate cases to 21% in severe cases. Patients on Medicaid, those with comorbid anxiety, and those flagged for suicide risk had lower odds of both response and remission. More frequent clinical contact was associated with better outcomes across all groups. (link)

⏺️ RCT found self-adaptive cognitive training outperforms fixed training in schizophrenia. A double-blind trial of 72 patients compared a programme that automatically adjusts its difficulty level to the user's performance against one that stays fixed. The self-adaptive group improved faster and more significantly on attention and processing speed, and also showed greater reductions in positive symptoms at week four. The active ingredient appears to be the continuous difficulty adjustment rather than the training content itself. (link)

⏺️ A meta-analysis confirmed intranasal esketamine combined with an oral antidepressant reduces treatment-resistant depression. Six RCTs covering 1,836 patients found significant improvements in response and remission, with rapid onset by day 2 and a 49% reduction in relapse risk during maintenance. (link)

⏺️ Pew Research found 6 in 10 Americans support banning social media for under-16s. Support crossed party and demographic lines, though young adults remained less supportive than older age groups. Support for social media policies targeting minors has grown since 2023. Despite the public support, many leading researchers argue that such bans will not be effective. (link)

⏺️ No differences found between intervention types and formats for people with antenatal depression. A meta-analysis of 115 RCTs found a moderate pooled effect size (SMD=0.5) for interventions targeting antenatal depression, but interestingly, no significant differences between intervention types or formats. (link)

⏺️ RCT found peer-led CBT group reduced depression and suicidal ideation. Researchers in South Korea randomised 118 adults with depression to either seven peer-led videoconferencing sessions combining CBT techniques with shared lived experience, or a CBT mobile app. Both reduced depression comparably, but only the peer-led group significantly reduced suicidal ideation, with 75% of participants achieving clinically meaningful improvement. Effects held at three-month follow-up. (link)

⏺️ ED visits for suspected suicide among 12 to 17-year-olds in the US decreased by 13% from 2021 to 2025. A CDC study in MMWR shared interesting data on ED visits for suspected suicides. While overall visit volumes have remained largely the same, some demographics have seen a reduction. (link)

⏺️ Scoping review found most crisis line innovations lack supporting evidence. Researchers reviewed 42 studies describing crisis line interventions beyond standard care. Only four found evidence supporting their intervention. Single-study support existed for suicide intervention training, structured safety planning, AI-assisted messaging, and a Beijing-developed risk assessment scale. The remaining interventions, including triage protocols, care coordination, and AI support tools, had not been formally evaluated. (link)

⏺️ Fingertip wearable detected stress and anxiety states with reasonable accuracy. Researchers tested a small device that shines light onto the fingertip to measure blood flow and tissue metabolism across 132 adults in 19 countries. A machine learning model used those signals to identify people with depression, anxiety, or stress symptoms, achieving an accuracy score with an AUC of 0.72. The sample was small and the findings are preliminary. (link)

⏺️ Psychology trainees show moderate-to-high acceptance of AI Precision Mental Health Technologies. A survey of 357 trainees found moderate-to-high acceptance and intention to use AI-PMHTs. Concerns about reliability and overdependence were the most common barriers. (link)

⏺️ A five-year study linked birdsong exposure to higher well-being, especially in people with mental health conditions. Researchers analysed 11,631 real-world assessments from 2,088 people via the Urban Mind smartphone app. Exposure to birdsong was associated with significantly higher momentary well-being (β = 1.34), with stronger effects in people with a mental health condition. The type of birdsong mattered too — acoustic variability drove the benefit more than birdsong presence alone. (link)

That’s all for this week.

Keep fighting the good fight!

Steve Duke

Founder of Hemingway

P.S. Welcome to Fun With Charts, with me, Steve Duke!

Fun with charts is our weekly competition to see who can guess the mental health chart. There was no winner of last week’s contest, so our prize rolls over. How exciting! So, what does the chart below represent? Just send your answer to this email to win our weekly Hemingway prize and be crowned chief mental health nerd (well, at least until next week).