Hi friend,

Onos Health raised $17M, NOCD launched their Intensive Outpatient Service and, of course, Sword is reportedly buying Headspace for up to $300M.

New research was announced by several industry players, including Two Chairs, Yuna and Brainsway. Anthropic launched a $5M grant programme, Meta settled their social media youth harm cases for $17.1B and New Zealand becomes the latest company to progress teen social media bans.

I enjoyed diving into several research papers this week too. The ODESSI trial was the big news of the week, shedding light on the efficacy of Open Dialogue in treating patients with major mental health challenges in the UK. There’s also a good paper on how AI models have improved in their handling of most crisis scenarios (but not all) over the last two years. We even have a paper on sugar and a separate paper on salt this week!

Let’s get into it!

If you find these updates valuable, consider joining more than 500 other mental health leaders who are Hemingway Pro members. As a Pro Member, you get access to deep industry insights, access to the Hemingway Slack group and invites to our member-only events. These include tactical sessions like How to take a digital health tool from pilot to full-scale adoption, with Trina Histon, as well as our private, in-person events around the world.

What happened in mental health last week?

Industry News:

⏺️ Sword Health is reportedly acquiring Headspace in an all-cash deal worth up to $300M. The deal would combine Sword's virtual musculoskeletal and behavioural health platform, including its AI-powered Mind and Dawn products, with Headspace's consumer app, virtual therapy, coaching and psychiatry services. Headspace had raised roughly $321M to date, and the reported valuation would represent a steep valuation drop from the $3B value set at its 2021 merger with Ginger. (link)

⏺️ Onos Health raised $17M. Onos is a clinical intelligence platform for health plans that uses AI to analyse therapy notes, intake assessments and utilisation histories to identify care gaps and quality issues. The Series A was led by Costanoa, with Flare Capital Partners and CVS Health Ventures also participating. The company says its platform has improved clinical standard adherence by 35% and cut behavioural health programme costs by more than 6% within a year. (link)

⏺️ NOCD launches its virtual intensive outpatient OCD programme. The programme delivers three hours of structured exposure and response prevention therapy a day, four days a week. It currently serves members in Colorado, Montana, Michigan and Texas with six more states on the waitlist. The company intends for the IOP to be available across the US. (link)

⏺️ Brightn and OpenLoop Health partnered to combine an AI-powered daily mental wellness app with nationwide licensed therapy. The partnership will package Brightn’s AI-based wellness app with access to OpenLoop's nationwide clinical network for employers and other health organisations. Brightn provides non-clinical daily journalling, mood tracking and AI support. Users who need clinical care will now be able to book a licensed therapist through OpenLoop through the Brightn app. (link)

⏺️ NFL and NFLPA launched a zero-cost behavioural health network for players. The NFL and NFL Players Association added a dedicated behavioural health network to the NFL Player Insurance Plan, expanding outpatient mental health coverage for current and former players and eligible dependants. The Cigna-administered network will cover every NFL market and provide personalised help finding providers and available appointments. It follows growing NFL efforts to expand mental health support for current and former players. (link)

⏺️ Two Chairs released research on time to response and remission in psychotherapy for depression. The study analysed 7,161 adults receiving psychotherapy for moderate-to-severe depression at Two Chairs. Among patients who achieved a clinical response (defined as at least a 50% reduction in PHQ-9 scores), the median time to response was six sessions, while those reaching remission did so in eight. The observational study had no comparison group, so it could not establish whether Two Chairs produced better outcomes than other forms of care. Findings from the study also supported “using complexity assessments for initial treatment planning and ongoing symptom monitoring for mid-treatment adjustments”. (link)

⏺️ Yuna released research showing users reported immediate improvements in mood and stress. The peer-reviewed study analysed 5,549 adults using Yuna, an AI-based mental health intervention, and found moderate within-session improvements in mood, d=0.55, and stress, d=0.56. Improvements between sessions were much smaller, and the retrospective study had no control group, so it could not establish that Yuna caused the changes. (link)

⏺️ BrainsWay released research linking Deep TMS to reduced suicidality in depression. The company highlighted a secondary analysis of a multisite randomised trial comparing accelerated TMS with once-daily Deep TMS for major depression. Suicidality declined with both treatment schedules, with accelerated treatment associated with earlier improvement. Suicidality was a secondary outcome, making the findings exploratory, and several authors reported ties to BrainsWay. (link)

⏺️ New Zealand introduced legislation to ban social media for under-16s. The government introduced its Online Safety Bill to Parliament, moving a policy it had been developing since 2025 into formal legislation. The bill would require tech platforms including Instagram, TikTok, Snapchat and Facebook to take reasonable steps to verify users are at least 16, using methods such as existing account data, facial age estimation and digital ID. It would also create broader child-safety obligations for online platforms, including some AI companion services, with companies facing penalties of up to 10% of global revenue for systemic non-compliance. Coalition partner, New Zealand First, said it would not support the legislation. Social media bans remain a hotly contested legislative topic in many countries. (link)

⏺️ Meta agreed to pay up to $17.1B to settle youth social media claims. Meta reached the landmark settlement during a federal trial brought by US states alleging that Facebook and Instagram were deliberately designed to encourage addictive use among children and contributed to harms to young people's mental health. Meta will pay about $12.7B to participating states over 10 years, with a further $5.3B contingent on TikTok and YouTube adopting similar child-safety measures. The agreement also requires changes to Meta's products, including limits on teenagers' daily use, overnight restrictions, stronger age verification and parental controls. Meta denied any wrongdoing. (link)

⏺️ Anthropic launched a $5M grant programme to fund independent research into how AI affects users' wellbeing. The grant programme will give researchers direct funding, access to Anthropic's models and technical support to build open-source evaluations and benchmarks, particularly for sensitive contexts such as mental health crises. Applications are due September 21st, with full proposals requested by October 5th. (link)

Join more than 500 other mental health leaders who are Hemingway Pro members. As a Pro Member, you get access to deep industry insights, access to our private Slack group and invites to our member-only events.

Research News:

⏺️ Long-awaited ODDESSI trial found no relapse benefit from Open Dialogue, but improvement in several secondary outcomes. Open Dialogue is a Finnish-developed model of crisis care that brings family and social networks into treatment and emphasises rapid support, shared decision-making and continuity with the same clinical team. Early non-randomised studies generated substantial international interest, but the approach had never been tested in a large randomised trial. The NIHR-funded ODDESSI trial was established to provide that evidence in the NHS. Recruitment began in 2019, and participants were followed for two years. Researchers studied 494 adults across 30 clusters in five English NHS trusts and found Open Dialogue did not significantly improve the primary outcome, time to first relapse after recovery, HR 0.95. However, patients had fewer psychiatric admissions and re-referrals to crisis or secondary mental health services and reported better recovery, quality of life and satisfaction. (link)

⏺️ An independent evaluation found leading AI chatbots have become much better at not encouraging suicide, but many still assist with suicide-related creative writing and other risky tasks. The independent nonprofit Transluce conducted what it described as the largest evaluation to date of how leading models respond to suicidal ideation, psychosis and mania. Researchers simulated more than 50,000 multi-turn conversations, covering 77 model variants released between May 2024 and July 2026, and assessed 14 behaviours. Newer models were substantially less likely to reinforce delusions or facilitate suicide and more likely to encourage human support than older models, although they still sometimes assisted with potentially harmful tasks such as writing farewell notes. The evaluation used simulated rather than real crisis conversations, although Transluce used anonymised information from real ChatGPT and Claude usage to make and validate its simulations. (link)

⏺️ A study found that estimates of how many people discuss mental health with AI chatbots can vary drastically depending on how "mental health use" is defined. There is growing interest in how often people turn to general-purpose AI for mental health support, but estimates depend heavily on what researchers count as a mental health conversation. RAND and Harvard researchers analysed 620,699 publicly available ChatGPT conversations and applied definitions ranging from broad mental health content to explicit psychological help-seeking. Just 0.21% met the strictest definition, while 4.90% contained mental health content under the broadest — a 23-fold difference. The researchers concluded that estimates of mental health AI use are highly sensitive to how it is defined, making consistent definitions important when measuring use and assessing safety. The study was based on analysis of publicly shared conversations, which may not represent how people use ChatGPT privately. (link)

⏺️ Research linked problematic short-video use to anxiety in Chinese adolescents. As evidenced by the New Zealand teen social media ban and the Meta settlement last week, the mental health effects of TikTok-style short-form video are an area of global concern. To shed light on this topic, researchers surveyed 3,261 adolescents and measured problematic use, anxiety, cognitive flexibility and exercise. Problematic short-video use was associated with higher anxiety, r=0.329 and lower cognitive flexibility, r=-0.286. Statistical modelling suggested cognitive flexibility mediated part of the association, but the cross-sectional design could not establish its direction or causality. The findings are consistent with a recent systematic review of 42 studies involving young people, which linked heavier short-video use with greater anxiety, depression and inattention. (link)

⏺️ Research found psilocybin reorganised brain activity to become more sensitive to context. Researchers gave 62 adults psilocybin and used fMRI and EEG while participants rested, meditated, listened to music and watched a film. Psilocybin reorganised brain activity so that it more strongly reflected participants’ immediate experiences, producing more distinct patterns during meditation, music, film watching and rest. Stronger context alignment was associated with more intense self-dissolving experiences and greater next-day changes in mindset. The mechanistic study did not test whether these brain changes improved mental health symptoms. (link)

⏺️ A UK Biobank study found children exposed to longer periods of sugar rationing in early life had lower rates of anxiety as adults. Researchers analysed 46,448 adults born in England between 1951 and 1956, comparing those exposed to different durations of wartime sugar rationing during their first 1,000 days of life. Those with the longest exposure to rationing had a 33% lower hazard of anxiety and a 29% lower hazard of depression compared with those never exposed. The anxiety finding remained robust after adjusting for later-life sugar intake, while the depression association weakened and became non-significant. The findings are observational and cannot establish that early sugar restriction directly causes better adult mental health. (link)

⏺️ Analysis found night-time PHQ-9 submissions had higher rates of severe symptoms. Digital screening means patients increasingly complete depression questionnaires outside clinic hours, raising questions about whether the timing of submission has any relevant clinical information. Epic Research analysed 616,683 PHQ-9s and found higher rates of severe symptoms in late-night submissions. Specifically, 10.1% reported thoughts of self-harm or being better off dead, compared with 4.2% at 6 am. Only around 2.5% of questionnaires were submitted overnight, so nearly 97% of all positive self-harm responses still occurred during the day. The vendor-published analysis was descriptive and could not reliably distinguish patient-entered questionnaires from those entered by clinicians. I’m not sure there’s actually that much to be learned here, but some of the time-series data in the report is interesting to consider. (link)

⏺️ Research found regular nature contact was linked to a smaller mental health gap between income groups. Researchers surveyed 1,433 adults in Sydney and Brisbane and found that lower-income participants generally reported worse mental health than higher-income participants. However, these differences in depression, anxiety and stress were smaller among people who visited nature at least once a week, with much of the association seen among lower- and middle-income groups. Spending at least two hours a week in nature was also associated with a smaller income gap in wellbeing. The study was cross-sectional, so it cannot show that nature contact caused these differences. (link)

⏺️ 1 in 10 South Korean children and adolescents estimated to have an anxiety disorder. Researchers assessed a nationally representative sample of 6,275 people aged 6 to 17 and estimated anxiety disorder prevalence at 10.2%. Anxiety was 36% more likely among girls, 83% more likely among those with disrupted sleep, 51% more likely with little physical activity and 74% more likely among those living in urban areas. Peer victimisation was associated with more than double the likelihood of anxiety, while parental divorce was associated with nearly triple. The observational study cannot establish causation. (link)

⏺️ New article identifies a salt-float protocol as a tractable, non-pharmacological method for inducing specific altered states of consciousness. Floatation-REST is a protocol that reduces external sensory input by having people float weightlessly in a dark, quiet pool of warm salt water. Previous research has suggested it may reduce anxiety. Researchers conducted a secondary analysis of a randomised feasibility trial involving 75 treatment-seeking adults with anxiety and depression, comparing six Floatation-REST sessions with a zero-gravity chair condition. Floatation increased awareness of internal bodily sensations and produced changes in perceived boundaries between the body, self and environment, an altered state the researchers termed "aquahenosis". The study suggests these altered states may help explain some of Floatation-REST's psychological effects, but it was a small, exploratory secondary analysis and was not designed to establish clinical efficacy. (link)

⏺️ Editorial called for more research on menopause-related cognition at work.Cognitive difficulties such as problems with memory, concentration and finding words are commonly reported during menopause, but receive less attention than other symptoms. Writing in Aging & Mental Health, researchers argued that these symptoms can affect women's confidence and ability to work, and called for better research, workplace support and clinical recognition. (link)

That’s all for this week.

Keep fighting the good fight!

Steve Duke

Founder of Hemingway

P.S. If you find these updates valuable, consider becoming a Hemingway Pro Member. As a Pro Member, you will get access to even deeper industry insights and will be invited to join our community of over 500 mental health leaders.