How Elggo’s clinical AI Is keeping the GCC’s young people safe

In an interview with HealthTechAsia, Mirna Mneimneh and Luma Makari, Co-Founders of Elggo, a UAE-based AI mental health platform built specifically for Arab youth, explain how they are constructing the region’s wellbeing infrastructure from the ground up, and why the Arab world doesn’t need adapted Western tools.

Elggo is described as Arabic-first and clinically informed. What does that mean in practice?

Arabic-first means the AI reads Arabic natively. It was trained on Arabic-language clinical and mental health datasets, not translated from English. That’s a meaningful technical distinction. Arabic emotional expression carries different weight and nuance than English.

Dialect, metaphor, code-switching between Arabic and English. A translated model misses all of it. Clinically informed means every tool on the platform is grounded in evidence-based frameworks like CBT, ACT, and DBT, validated against Arab and GCC populations specifically.

Our assessment instruments are clinically validated with the populations we serve. The AI scores journal entries against 15 wellbeing metrics in real time, detecting mood complexity, cognitive distortions, and passive distress signals. This isn’t a general-purpose chatbot with a mental health layer on top. It’s a purpose-built wellbeing model with clinical-grade capability embedded at the architecture level.

How does AI-powered journalling work, and what clinical guardrails are in place?

Students can journal freely or use guided prompts, in Arabic or English. The AI processes each entry in real time, analyzing emotional tone, distress signals, and thinking patterns.

It builds a continuous emotional baseline for each student over time, so it’s not reacting to a single entry in isolation. It’s tracking patterns and detecting drift. The guardrails are structural, not optional. Clinical thresholds are hard-coded into the system. Schools cannot adjust them. When a distress signal crosses a threshold, it routes immediately to the school counselor.

No raw journal content is ever visible to staff. Only risk flags and sentiment-level data surface to authorized roles. The AI does not diagnose, does not give clinical advice, and does not present itself as a practitioner. It’s a detection and reflection tool with escalation pathways built into its core.

How do you think about the boundary between AI-assisted support and clinical intervention, and how is that enforced technically?

The boundary is defined by what the AI is allowed to do and what it isn’t. The AI supports reflection, emotional expression, and early awareness. The moment a signal suggests clinical-level concern, the system hands off to a human. That handoff is enforced at the model level.

Our wellbeing scoring engine analyses text input against 15 metrics and generates 30-day behavioral trend forecasts. Deviations from a student’s established baseline trigger flags, not absolute thresholds alone. This means the system catches subtle drift, not just acute crisis language.

Escalation contacts and pathways are configured per deployment before any student onboards. In school settings, flags route to the counselor. The clinical thresholds that trigger escalation are hard-coded and not configurable by end users.

There is no grey area in the system architecture about when AI stops and humans start.

What does responsible AI deployment look like in your context, and how do you approach data privacy for minors?

Young people are using AI regardless. The question is whether they’re using it inside a safe, governed environment or outside one. Elggo is a walled garden. No student data leaves the platform. No third-party access. User data is never sold or shared.

We’re GDPR-aligned, UAE PDPL-compliant, and FERPA and HIPAA-aligned. Data sovereignty is enforced: servers are hosted locally in each implementing country. Zero personally identifiable information is accessible at the admin level.

The AI is culturally calibrated for the Arab world, designed around how young people in this region actually communicate. And every deployment context has its own safeguarding configuration set up before launch. Responsible deployment for us means the safety architecture exists before the first student logs in, not after.

How do you self-regulate in the absence of formal AI-in-mental-health regulation in the region?

We don’t wait for regulation to tell us what safe looks like. We apply the strictest standards available globally and layer regional requirements on top. GDPR, UAE PDPL, FERPA, HIPAA alignment.

Clinical thresholds hard-coded, not configurable. Escalation pathways mandatory before deployment. No clinical data collected in school or community settings. In healthcare deployment contexts, clinical governance frameworks are agreed with the partner before launch and a clinical toggle tightens the AI’s thresholds and output accordingly.

We also built in structural accountability: no raw student content visible to faculty, tiered permissions across five user roles with strict data access boundaries, and walled-garden architecture with no external data sharing. We treat self-regulation as a floor, not a ceiling.

Your WHO-5 outcomes data is a strong proof point. How is that measured?

WHO-5 is administered as part of student onboarding to establish a baseline wellbeing score, then measured again at programme completion. It’s a clinically validated five-item scale widely used in global health research. We run it across every school deployment.

Across our schools, we’ve measured a 30% improvement in WHO-5 wellbeing scores. Beyond WHO-5, we measure session-level knowledge gains through pre and post assessments on every lesson, track journal engagement patterns, and cross-reference onboarding data with ongoing check-in and assessment responses for consistency.

The system also flags rapid-click responses as unreliable and excludes them from analytics. We don’t rely on engagement metrics alone. We measure actual wellbeing outcomes, knowledge shift, and behavioral indicators because that’s what schools, governments, and clinical partners need to see.

You’ve reached 17,000+ young people across six countries. What does the next phase look like?

System-wide adoption. Moving from school-by-school deployment to nationwide infrastructure. That means integration with ministries of education, health systems, and government platforms. We’re already in conversation with partners across the GCC on exactly this.

On the technology side, the next phase includes expanding our Arabic NLP capabilities through regional research partnerships, scaling the healthcare deployment context with clinical-toggle functionality for licensed settings, and deepening the parent-facing layer so families have real tools, not just awareness.

We’re also building government-level analytics: anonymized, population-level wellbeing dashboards aligned to national strategy frameworks. The goal is that Elggo becomes the wellbeing infrastructure layer across education and health systems in the region. Not an add-on. The standard.