AI SAFETY & LIMITATION POLICIES

AI USE, SAFETY & LIMITATION DISCLAIMER

Applies To: Neurolens, all AI-generated outputs, preliminary assessments, screening insights.
Issued by: SAHCHI HEARING AND SPEECH SOLUTIONS PRIVATE LIMITED (“Gabify”)

1. PURPOSE OF THIS DISCLAIMER

This section explains:

• What AI in Neurolens can do
  
• What AI cannot do
  
• Limitations and risks
  
• User responsibilities
  
• Boundaries of liability
  

This is required under AI ethics guidelines and medical-technology compliance worldwide.

2. NEUROLENS IS NOT A DIAGNOSTIC TOOL

Neurolens:

• Does not diagnose Autism, ADHD, or any developmental condition
  
• Does not replace clinical evaluation, observation, or standardized assessments
  
• Should not be used as the sole basis for intervention, therapy, or certification decisions
  
• Provides preliminary insights only, intended for professional interpretation
  

All AI outputs require human review and validation.

3. LIMITATIONS OF AI-BASED INSIGHTS

Users understand and agree:

3.1 AI May Produce Inaccurate or Incomplete Results

• AI may misinterpret behaviors
  
• AI outputs depend on quality of input data
  
• Environmental noise may affect video/audio interpretation
  

3.2 AI Cannot Understand Nuance Like a Clinician

• Cultural variations
  
• Contextual cues
  
• Emotional and social subtleties
  
• Family and environmental factors
  

3.3 AI Cannot Replace Standardized Tools

Such as:

• CARS
  
• M-CHAT
  
• ADOS
  
• Clinical neurological evaluation
  

AI is supportive, not definitive.

4. NO AUTOMATED DECISION MAKING

Neurolens does not make independent determinations.

It is designed to:

• Support clinician thought processes
  
• Highlight observable patterns
  
• Reduce documentation workload
  
• Improve workflow consistency
  

Clinicians retain full decision-making authority.

5. AI IS PROBABILISTIC, NOT CERTAIN

AI outputs may include:

• Suggested observations
  
• Pattern indications
  
• Behavioral markers
  
• Probabilistic insights
  

These do not equate to clinical conclusions.

6. NO LEGAL, MEDICAL, OR CERTIFICATION USE

Neurolens outputs must not be used as:

• Medical certificates
  
• Disability certificates
  
• Legal evidence
  
• School placement decisions without evaluation
  
• Therapy continuation/discontinuation decisions
  

Gabify is not responsible for misuse.

7. USER AGREES TO ASSUME RESPONSIBILITY

By using Neurolens, the user acknowledges:

• AI is fallible
  
• Human oversight is mandatory
  
• AI outputs must be contextualized clinically
  
• Gabify does not guarantee accuracy
  

8. LIMITATION OF LIABILITY

Gabify is not liable for:

• Misinterpretation of AI insights
  
• Clinical decisions made by users
  
• Use of AI outputs outside intended scope
  
• Incorrect assessments due to faulty input data
  
• Improper use by unqualified persons
  

9. CHANGES TO AI MODELS

Gabify may:

• Update algorithms
  
• Improve accuracy
  
• Retrain models
  
• Add new screening capabilities
  

without prior notice, as part of continuous improvement.

CLINICAL RESPONSIBILITY & SUPERVISION POLICY

 Applies to ALL clinical users (SLP, psychologist, therapist, paediatrician, special educator).

1. PURPOSE

This policy clarifies:

• Clinical responsibilities
  
• Supervision requirements
  
• Appropriate use of Neurolens
  
• Legal expectations from professionals
  

It ensures clinical safety and compliance.

2. PROFESSIONAL SUPERVISION MANDATE

Neurolens must be used only under supervision of a qualified professional, such as:

• Speech-language pathologist
  
• Psychologist
  
• Paediatrician
  
• Behaviour therapist
  
• Special educator
  

The supervisor is responsible for:

• Evaluating AI outputs
  
• Making final decisions
  
• Providing clinical context
  
• Ensuring ethical practices
  

3. RESPONSIBILITIES OF THE CLINICIAN

Clinicians must:

3.1 Review AI Outputs Carefully

Confirm, modify, or reject AI-generated insights.

3.2 Conduct Independent Observations

Neurolens is only one component of the evaluation.

3.3 Obtain Proper Consent

Including parental consent for child assessments.

3.4 Use Professional Judgment

Clinical reasoning supersedes machine interpretation.

3.5 Ensure Ethical Use of AI

Especially in cases involving:

• Vulnerable children
  
• Non-verbal individuals
  
• Neurodiverse populations
  
• Sensitive family situations
  

4. SCOPE OF PRACTICE LIMITATIONS

Neurolens may not be used to:

• Diagnose Autism
  
• Diagnose ADHD
  
• Provide medical approval
  
• classNameify severity levels
  
• Certify disability
  
• Recommend medication
  

These require professional evaluation.

5. RESPONSIBILITY OF INSTITUTIONS

Institutions must ensure:

• Only licensed professionals use the platform
  
• Users receive training
  
• Staff follow ethical and clinical protocols
  
• Access controls are implemented
  
• Consent obligations are fulfilled
  

6. NON-CLINICAL STAFF RESTRICTIONS

Administrative staff may access:

• Scheduling tools
  
• Usage dashboards
  
• Billing interfaces
  

They may not access:

• Clinical reports
  
• Child audio/video
  
• AI screening outputs
  
• Assessment histories
  

unless explicitly authorized.

7. PROHIBITION OF UNSUPERVISED USE

Neurolens must not be used by:

• Parents
  
• Students
  
• Volunteers
  
• Untrained staff
  
• Anyone not clinically certified
  

Gabify may terminate access for misuse.

8. LIABILITY & DISCLAIMER

Clinicians acknowledge:

• Final decisions rest with them
  
• AI does not replace diagnostic frameworks
  
• Gabify is not liable for clinical errors
  

ALGORITHMIC FAIRNESS, BIAS & QUALITY POLICY

 Applies to AI models used in Neurolens for screening, interpretation, and insight generation.

1. PURPOSE

This policy outlines how Gabify ensures:

• Ethical use of AI
  
• Mitigation of bias
  
• Transparent functioning
  
• Model performance monitoring
  
• Fair treatment of diverse populations
  

2. COMMITMENT TO FAIRNESS

Gabify commits to:

• Avoid discrimination
  
• Prevent bias against any demographic
  
• Support inclusive datasets
  
• Evaluate edge cases
  
• Maintain ethical standards
  

Neurolens AI models are trained using:

• Diverse samples
  
• Multiple age groups
  
• Multilingual speech patterns
  
• Varied socio-cultural backgrounds
  

3. IDENTIFIED SOURCES OF BIAS

Potential bias may arise from:

• Under-representation of certain populations
  
• Variations in recording environment
  
• Cultural behavioral differences
  
• Socio-linguistic variation
  

Gabify continually refines models to address these.

4. MEASURES TO MITIGATE BIAS

Gabify implements:

4.1 Human-in-the-Loop (HITL)

Clinicians always interpret results.

4.2 Diverse Data Training

Only de-identified data with consent.

4.3 Model Evaluation Metrics

Accuracy
Precision
Recall
False positives/negatives

4.4 Regular Audits

Bias tests
Output consistency checks

4.5 Clear User Disclaimers

AI limitations are disclosed prominently.

5. TRANSPARENCY OF AI BEHAVIOR

Gabify ensures that:

• AI suggestions are explainable
  
• Users understand why certain patterns were flagged
  
• Reports indicate uncertainty when relevant
  

6. CONTINUOUS MODEL IMPROVEMENT

Neurolens AI models evolve through:

• Performance monitoring
  
• Feedback from clinicians
  
• De-identified consented data
  
• Error analysis
  
• Feature updates
  

7. USER RESPONSIBILITY TO REPORT ERRORS

Clinicians must report:

• Incorrect insights
  
• Bias indicators
  
• MisclassNameifications
  
• Harmful outputs
  

to info@gabify.life

Gabify investigates all such reports.

8. NO USE FOR AUTOMATED PROFILING

Neurolens will not:

• Automatically classNameify children
  
• Categorize them into risk labels without human oversight
  
• Make irreversible recommendations
  

AI outputs are always suggestive, not determinative.