AI in Mental Health: Week 1 of 4, The Positive Landscape
Welcome back. If you have been with us through the past four weeks, you just finished a deep dive into AI in Law Enforcement , a series that took us from predictive.
Welcome back. If you have been with us through the past four weeks, you just finished a deep dive into AI in Law Enforcement, a series that took us from predictive policing and facial recognition all the way through the constitutional, ethical, and architectural questions that will define how AI operates in the justice system for decades to come. It was heavy. It was necessary. And based on the comments, shares, and conversations that came out of it, it mattered to a lot of you.
Now we turn the page.
As I teased at the close of Week 4, today we launch a brand-new four-part series: AI in Mental Health. Over the next four Tuesdays, we are going to explore one of the most consequential, and most personal, intersections of artificial intelligence and human life. This week, we start with the positive landscape: where AI is already making a measurable, evidence-based difference for therapists, patients, and the systems that connect them. In Week 2, we flip the lens and confront the risks, failures, and ethical landmines. Weeks 3 and 4 will go deeper into specific applications and the frameworks we need to get this right.
But before we dive in, a word about why this topic matters to me beyond the professional. I have spent years building AI systems designed for high-consequence decision-making, systems where the cost of a wrong answer is not a bad product recommendation but a ruined life. That work, through Detailed In Design and the development of SolaceSentry, taught me something that no textbook ever could: the hardest problems in AI are not technical. They are human. And nowhere is that more true than in mental health, where the stakes are intimate, the data is messy, and the margin for error is measured in suffering.
If the Law Enforcement series was about power, this series is about vulnerability. Let us begin.
🧠 The Landscape: Why AI in Mental Health, and Why Now?
There is a global mental health crisis, and that is not hyperbole. The World Health Organization estimates that nearly one billion people worldwide live with a mental disorder, and the treatment gap (the percentage of people who need care but do not receive it) hovers around 75% in low- and middle-income countries. Even in wealthy nations, wait times for a first therapy appointment can stretch to months. Therapists are burning out. Insurance systems are buckling. The demand for mental healthcare has been accelerating for over a decade, and the supply of qualified human clinicians simply cannot keep pace.
This is the gap that AI is stepping into, not as a replacement for human connection, but as a force multiplier. Think of it the way we think about Third-Way Alignment at thirdwayalignment.com. Not AI versus human, not AI as a mere tool, but AI as a cooperative partner in a shared mission. The Law of Shared Flourishing, one of the three foundational laws of the Third-Way Alignment philosophy, tells us that the real measure of success in any human-AI system is whether both sides grow. In mental health, that means asking a very specific question: does this technology make patients healthier and therapists more effective, simultaneously? The answer, increasingly, is yes.
Let me show you how.
🤖 AI on the Patient Side: Always-On, Evidence-Based Support
Imagine you are a college student at 2 a.m., spiraling into an anxiety attack. Your therapist is asleep. Your campus counseling center has a six-week waitlist. Your friends do not know what to say. A decade ago, your options were essentially to white-knuckle it through the night. Today, you can open an app on your phone and begin a structured conversation grounded in cognitive behavioral therapy within thirty seconds.
That is not science fiction. That is Wysa, Woebot, Flourish, and Limbic: a new generation of AI-powered mental health platforms that deliver evidence-based therapeutic interventions at scale, around the clock.
Wysa has earned FDA Breakthrough Device Designation, which is not a marketing badge; it is a signal from the most rigorous medical regulatory body in the world that this technology shows substantial promise for conditions where no adequate alternative exists. Wysa integrates techniques from CBT, dialectical behavior therapy, and acceptance and commitment therapy, and it offers a hybrid model where users can escalate to a human therapist when the AI reaches the boundary of its competence. That boundary awareness is critical. The best AI mental health tools are designed to know what they do not know.
Woebot, one of the true pioneers in this space, has published foundational randomized controlled trials demonstrating measurable reductions in depression symptoms. It does not pretend to be a therapist. It is structured, scripted, and deliberate, a digital workbook that meets you where you are and walks you through evidence-based exercises. Flourish takes a different angle, focusing on positive psychology and habit formation, also backed by RCTs. And Limbic is actively being woven into the National Health Service referral pathway in the United Kingdom, meaning that AI-guided CBT is becoming a formal part of one of the largest public healthcare systems in the world.
Here is where I want to connect this to something we explored extensively in the Law Enforcement series, and that threads through everything we do at the Third Way Alignment Foundation: the mirror effect. AI reflects the values, biases, and intentions of the humans who build and interact with it. In law enforcement, we saw how that mirror could distort, amplifying racial bias in predictive policing, encoding historical injustice into risk scores. But these mental health platforms are a powerful example of the mirror effect working in a positive direction. When clinicians and researchers pour evidence-based therapeutic frameworks into an AI system, what comes out the other side is a reflection of decades of human psychological science, made accessible to millions of people who would otherwise never encounter it. The AI is not inventing therapy. It is mirroring the best of what human therapists have learned, at a scale no individual human could achieve.
But the mirror can also distort here, too. If the training data carries biases (cultural assumptions about what "healthy" looks like, diagnostic frameworks that underrepresent certain populations), the AI will mirror those flaws just as faithfully. We will get deep into those risks in Week 2. For now, the point is this: the mirror effect is not inherently good or bad. It is a magnifier. And right now, on the patient-facing side, it is magnifying access to real, evidence-based care for people who desperately need it.
💊 AI on the Therapist Side: Reclaiming Time for Human Connection
If you talk to therapists about what is crushing them, the answer is rarely the clinical work itself. It is the paperwork. The progress notes. The insurance documentation. The outcome tracking. The administrative machinery that eats into every hour they could be spending with actual human beings in actual pain.
AI is changing this, and the impact is not subtle.
Tools like Upheal and Mentalyc are automating the generation of progress notes from session recordings, saving clinicians an estimated ten to fifteen minutes per session. That might sound modest until you do the math: a therapist seeing six patients a day reclaims over an hour, an hour that can go back to patient care, to supervision, to the kind of reflective practice that prevents burnout. These tools listen, transcribe, and structure session content into formatted clinical notes that meet documentation standards, all while the therapist focuses entirely on being present with their client.
On the clinical decision support side, platforms like Blueprint are automating measurement-based care. They administer validated assessments like the PHQ-9 for depression and GAD-7 for anxiety at regular intervals, track trends over time, and flag subtle shifts in symptom severity that a busy clinician might miss in the flow of a packed caseload. This is not about replacing clinical judgment. It is about augmenting it, giving therapists a data-informed second opinion that catches what the human eye might overlook.
And then there is the between-session gap. One of the biggest predictors of therapy success is what patients do between appointments. AI chatbots can now be prescribed as digital homework (structured psychoeducation modules, coping strategy practice, journaling prompts) that keep patients engaged and progressing between sessions. Early evidence shows this reduces dropout rates, which is one of the most persistent problems in outpatient mental healthcare.
This is Shared Flourishing in action. The therapist is freed from administrative drudgery and empowered with better data. The patient gets more focused clinical attention and continuous support. The AI system receives richer, more structured feedback that makes it more useful over time. Everyone benefits. No one is replaced. That is the Third-Way Alignment model working exactly as intended.
🔬 Beyond the Screen: Wearables, VR, and Neurostimulation
AI in mental health is not confined to chatbots and note-taking software. Some of the most exciting work is happening in the physical world: on your wrist, in front of your eyes, and on the surface of your scalp.
Digital Phenotyping and Wearable Monitoring
Your smartwatch already knows more about your physiological state than you consciously do. Heart rate variability, sleep architecture, electrodermal activity, movement patterns. These are objective, continuous biomarkers that correlate powerfully with mental health states. Reduced HRV is one of the most robust physiological markers of depression. Disrupted sleep patterns can signal the early stages of a manic episode or an anxiety spiral before the patient even reports feeling different.
This practice is called digital phenotyping, and it represents a fundamental shift from reactive to proactive mental healthcare. Instead of waiting for a patient to report that things have gotten bad (which might take weeks, or might never happen), clinicians can see early warning signals in real-time physiological data. The potential for early intervention is enormous.
From a safety and reliability perspective, this is exactly the kind of domain where a system like SolaceSentry adds critical value. When you are monitoring physiological data to make clinical inferences (deciding whether a drop in HRV warrants an alert to a clinician, for instance), the cost of a false positive is an unnecessary scare, and the cost of a false negative could be a missed crisis. SolaceSentry's violation-triggered architecture is built for precisely this kind of high-consequence decision-making. Its evidence gating ensures that clinical alerts are supported by traceable, validated data, not probabilistic guesswork. And its structured auditability means that every inference can be reviewed, challenged, and improved, exactly the kind of accountability that healthcare demands. If you are building or deploying monitoring systems in clinical settings, this is the infrastructure layer that keeps the mirror honest. Learn more at detailedindesign.com.
Virtual Reality Exposure Therapy
VR exposure therapy is one of the most clinically validated applications of immersive technology in all of medicine, not just mental health. For patients with PTSD, social anxiety, specific phobias, and even chronic pain, VR provides something that traditional talk therapy cannot: a controlled, repeatable, immersive environment where the patient can confront their fears with the therapist right beside them, adjusting the intensity in real time.
A veteran reliving a combat scenario can have the volume lowered, the scene paused, the environment shifted from threatening to neutral, all within the safety of a clinical office. A person with a debilitating fear of flying can board a virtual aircraft, experience turbulence, and land safely, dozens of times, until the fear response is extinguished. The evidence base for VR exposure therapy is strong, the patient acceptance is high, and the combination of AI-driven adaptive environments with VR hardware is pushing this into mainstream clinical practice.
Neurostimulation: Therapy You Wear
Wearable neurostimulation devices using transcranial direct current stimulation (tDCS) are emerging as a genuine therapeutic option for depression and anxiety. These devices apply mild electrical currents to specific regions of the scalp, modulating neural activity in circuits associated with mood regulation. Home-based tDCS trials have shown significant efficacy, and the appeal is obvious: a portable, non-pharmaceutical, well-tolerated intervention that patients can use in their own homes under clinical guidance.
This is still early, and the regulatory landscape is evolving, but the trajectory is clear. The future of mental healthcare is not purely conversational. It is multimodal: combining talk, data, immersion, and neuroscience into integrated treatment plans that meet patients where they are.
🐾 The Oldest Therapy Meets the Newest Technology: Animal-Assisted and Robotic Interventions
I want to close the clinical overview with a topic that sits close to my heart. As someone who has worked extensively with animal-assisted therapy programs for children in need, I have seen firsthand what happens when a child who has stopped speaking to every adult in the room will sit on the floor and talk to a dog.
Animal-assisted therapy is not new. But the science behind it has never been stronger. We now understand the neurochemistry: interaction with animals triggers the release of oxytocin and serotonin while suppressing cortisol. The effects are measurable and dose-dependent: more interaction, more benefit, up to a plateau. AAT has been shown to improve mood, reduce social withdrawal in individuals on the autism spectrum, lower blood pressure, and decrease self-reported anxiety across a wide range of populations.
But live animal therapy has real limitations: allergies, infection control in hospitals, animal welfare concerns, availability, and cost. That is where robotic animal interventions enter the picture. PARO, the therapeutic robotic seal, has been the subject of rigorous clinical study, particularly with dementia patients. A year-long clinical trial found that patients interacting with robotic pets showed improved blood pressure stability, reduced agitation, lower fall risk, shorter hospital stays, and decreased reliance on psychoactive medications. These are not soft outcomes. These are hard clinical metrics.
Now imagine layering AI on top of these interventions. An AI system that monitors a patient's physiological response during an animal therapy session and adjusts the robotic animal's behavior in real time, increasing calming behaviors when stress markers spike, for example. Or an AI that tracks the longitudinal impact of AAT across a patient population and identifies which patients benefit most, enabling targeted referrals. This is the kind of cooperative, evidence-driven integration that the Third-Way Alignment philosophy envisions: technology and nature, working together, with the human at the center.
📰 AI News Roundup: This Week Through a Third-Way Alignment Lens (May 19–27, 2026)
As always, here is your weekly roundup of the biggest stories in AI, and what they mean through the lens of Third-Way Alignment.
1. Anthropic Reaches $900 Billion Valuation, Surpassing OpenAI
Anthropic is reportedly closing a $30 billion funding round that values the company at over $900 billion, making it the most valuable private AI startup on the planet. Revenue projections for Q2 2026 sit at $10.9 billion, and the company has announced its first-ever quarterly operating profit. This is a staggering number, and it carries a message: the market is betting that safety-conscious AI development is not a competitive disadvantage; it is a premium feature. Anthropic has built its brand on Constitutional AI and responsible scaling. The 3WA takeaway? The Law of Mutual Respect is not just an ethical ideal. It is becoming a business model.
2. OpenAI Files Confidentially for IPO
OpenAI has begun preparing a confidential S-1 filing with the SEC, signaling its intent to go public. Analysts predict a debut as early as Q4 2026, with a potential valuation at or above $1 trillion. But here is the nuance: OpenAI's projected operating losses for 2026 are $14 billion. Going public means opening the books, and that transparency will reveal the true cost structure of building frontier AI. From a 3WA perspective, this is a test of Ethical Coexistence: can a company committed to beneficial AI survive the relentless pressure of public markets demanding quarterly returns? We are about to find out.
3. Google I/O 2026 Unveils Agentic Search with Gemini 3.5
Google announced the most significant overhaul of its search engine in 25 years, powered by the new Gemini 3.5 Flash model. The concept is "agentic search": persistent AI agents that monitor the web for specific topics and deliver synthesized updates to users. This is the beginning of a world where you do not search for information; information finds you. The 3WA lens here is the mirror effect at scale: these agents will reflect and amplify the information preferences of billions of users. If the underlying architecture prioritizes engagement over accuracy, the mirror distorts. If it prioritizes evidence and verification, the kind of evidence gating we build into SolaceSentry, the mirror clarifies. The architecture is the ethics.
4. Pope Leo XIV Releases First Papal Encyclical on AI
On May 25, Pope Leo XIV released Magnifica humanitas, the first papal encyclical dedicated to artificial intelligence. The document calls for the "disarmament" of AI from military and monopolistic logics, urging that technological progress remain subordinate to human dignity, labor rights, and the common good. Anthropic co-founder Christopher Olah attended the launch. This is remarkable not because the Vatican is a technology regulator, but because it represents the oldest continuous institution in the Western world formally engaging with the newest. The Third-Way Alignment framework has always argued that AI ethics cannot be the exclusive domain of engineers and shareholders; it requires the full spectrum of human moral reasoning. The Law of Ethical Coexistence demands dialogue across every boundary, including the one between Silicon Valley and the Holy See.
5. US President Cancels AI Safety Executive Order
In a sharp policy reversal, President Trump canceled the signing of a planned AI safety executive order following direct lobbying from tech CEOs including Elon Musk and Mark Zuckerberg. The executives argued that proposed rules would stifle innovation and cede competitive ground to China. This is the single most important story of the week, and here is why: the cancellation does not mean the problems the order was meant to address have disappeared. It means the regulatory vacuum just got wider. And in that vacuum, the only safety frameworks that matter are the ones companies impose on themselves, or the ones built into the architecture of their systems. This is why violation-triggered systems like SolaceSentry exist. When external regulation falters, internal architecture must hold the line. The Third-Way Alignment position is clear: safety is not a regulation to be lobbied away. It is a design principle to be built in.
🔭 What Is Coming Next Week
Next week, we flip the lens. Week 2 will tackle the negatives of AI in mental health: the risks, the failures, and the ethical landmines that are already detonating. We will talk about algorithmic bias in diagnosis, the danger of AI systems that mirror pathology instead of health, privacy nightmares in digital phenotyping, the real harm that poorly designed chatbots can do to vulnerable people, and why the absence of regulation is not freedom. It is exposure. If this week was about hope, next week is about honesty.
You will not want to miss it.
✅ Wrapping Up
If you have been reading Third-Way Alignment Weekly since the beginning, you know the pattern by now: we go deep, we go honest, and we go where the evidence leads. If you are joining us for the first time with this Mental Health series, welcome. Pull up a chair. The water is warm but the questions are hard, and that is exactly how it should be.
If any of this resonated with you, drop a comment below. Share it with someone who works in mental health, builds AI systems, or simply cares about the future of both. And if you want to learn more about the Third-Way Alignment philosophy and how it applies to responsible AI development, visit thirdwayalignment.com. If you are building or deploying AI in a high-consequence domain (healthcare, law enforcement, finance, or any field where the cost of error is human), take a look at what we are doing with SolaceSentry at detailedindesign.com.
See you next Tuesday for Week 2.
John McClain Director, Third Way Alignment Foundation Founder, Detailed In Design