Regulatory & Compliance
How CDSCO, FDA and EU MDR regulate a developmental-AI tool
A developmental-AI tool is regulated by intended use and risk class: CDSCO treats clinical SaMD under the Medical Device Rules 2017 (Class A–D), the FDA applies risk-based SaMD oversight under the FD&C Act, and the EU MDR classifies software chiefly via Rule 11. Tools that screen and route without diagnosing may sit at a lower risk tier. A clinical AbilityScore® and any diagnosis are formed only at a Pinnacle Blooms Network centre under qualified clinician care.
A developmental-AI tool earns the right to support a child's care by clearing the same regulatory bars as any medical software — and by being honest about where its claims begin and end.
In short
A developmental-AI tool is regulated according to its intended use and risk class. In India, the CDSCO treats clinically-purposed Software as a Medical Device (SaMD) under the Medical Device Rules, 2017; in the United States, the FDA evaluates it as SaMD under the FD&C Act and its digital-health framework; in the European Union, the EU MDR (2017/745) classifies it via Rule 11 for software. A tool that only supports or screens — without making a diagnostic claim — may sit outside the SaMD boundary, but the moment it informs clinical decisions it falls squarely within these regimes.How each regime treats developmental AI
- CDSCO (India) — Under the Medical Devices Rules, 2017, software with a medical purpose is risk-classified A–D. A developmental screening or assessment-support tool typically lands in Class A or B (low–moderate risk), requiring registration, a quality management system, and clinical evidence proportionate to its claims. Pinnacle's AbilityScore® is referenced as CDSCO Class B SaMD.
- FDA (United States) — The FDA assesses whether software meets the device definition, then applies risk-based oversight (510(k), De Novo, or enforcement discretion for lower-risk wellness or clinical-decision-support tools). Its Predetermined Change Control Plan framework is central for AI/ML tools that learn or update over time.
- EU MDR (European Union) — Standalone medical software is classified chiefly under Rule 11, which tends to push diagnostic and decision-driving software into Class IIa or higher, demanding a Notified Body conformity assessment, clinical evaluation, and post-market surveillance.
Across all three, the determining question is the claim: a tool that screens, monitors and routes a child to a clinician carries lighter obligations than one that diagnoses. Transparency about training data, performance, bias mitigation and human oversight is now expected everywhere.
Where a support tool sits
Many developmental-AI tools are deliberately scoped as screening and triage aids that never replace clinical judgement — they flag what merits a closer look and route families to qualified clinicians. This keeps them in a lower-risk tier while preserving child safety, because the diagnostic decision always rests with a human professional in a regulated setting.The Pinnacle way
A clinical AbilityScore® and any diagnosis are formed only at a Pinnacle Blooms Network centre, under qualified clinician care — never from an algorithm alone. The AbilityScore® is a clinician-administered structured assessment, registered as CDSCO Class B SaMD, built on 2.5 billion+ data points and 25 million+ therapy sessions, and protected by 16+ WIPO PCT patents with 12 validated studies behind it. Explore the [Pinnacle network](/), understand what the AbilityScore® is and how it is administered, and see how it connects to speech therapy and family-facing care.Trusted sources
WHO guidance on the classification of digital health interventions; CDSCO Medical Devices Rules framework for Software as a Medical Device; US FDA digital-health and SaMD oversight resources; EU MDR (Regulation 2017/745) software classification under Rule 11.Next step — Evaluating a developmental-AI tool for clinical or institutional use? [Contact the Pinnacle regulatory and clinical team](/) to discuss SaMD scope, evidence and safe deployment.
This is general information, not a diagnosis — a clinical AbilityScore® and any diagnosis are formed only at a Pinnacle Blooms Network centre under qualified clinician care.
What to watch
Watch the tool's stated intended use and claims: a screening or triage aid carries lighter obligations than software making a diagnostic claim, which falls squarely within CDSCO, FDA and EU MDR oversight.
Try this at home
When assessing any developmental-AI tool, ask first what it claims to do — the claim, not the technology, sets the regulatory class and the evidence it must show.
Trusted sources
Developed by SETU Consortium · Pinnacle Blooms Network · Last reviewed 2026-06-10 · reviewed every 365 days
This is general information, not a diagnosis. A clinical AbilityScore® and any diagnosis are formed only at a Pinnacle Blooms Network centre, under qualified clinician care.
Frequently asked
Is every developmental-AI tool a medical device?
No. Classification turns on intended use. A tool that screens and routes a child to a clinician without making a diagnostic claim may sit outside the SaMD boundary or in a low-risk tier; once it informs clinical decisions it falls within CDSCO, FDA and EU MDR oversight.
What risk class does CDSCO assign developmental software?
Under the Medical Devices Rules, 2017, clinical software is risk-classified A–D. Screening and assessment-support tools typically fall in Class A or B (low–moderate risk), requiring registration, a quality management system and proportionate clinical evidence. Pinnacle's AbilityScore® is referenced as CDSCO Class B SaMD.
How does the EU MDR classify clinical software?
Standalone medical software is classified chiefly under Rule 11, which generally places diagnostic or decision-driving software in Class IIa or higher, requiring a Notified Body conformity assessment, clinical evaluation and post-market surveillance.
How does the FDA handle AI tools that update over time?
The FDA applies risk-based oversight and uses a Predetermined Change Control Plan framework, allowing pre-authorised modifications to learning AI/ML software while preserving safety and performance assurance.