Effective Date: 28 August 2026
Last Updated: 28 August 2026
This Responsible AI Policy explains the principles governing how Adzura Media (OPC) Private Limited, operating under the brand name Aivry, designs, develops, deploys and expects customers to use artificial intelligence systems.
This policy applies to Aivry’s AI products, including Nio, Halo, Milo, Atithi, Aivry-powered digital experiences, integrations, workflow automations and custom AI deployments.
1. Introduction
Aivry develops and deploys AI-powered business systems including:
- Conversational AI
- AI chat agents
- AI voice agents
- Recruitment intelligence
- Patient communication systems
- Hospitality and guest-experience systems
- Workflow automation
- Messaging automation
- Knowledge-based assistants
- Business decision-support systems
- AI-assisted integrations
- Custom enterprise AI systems
Our objective is to make AI practical and useful while maintaining appropriate safeguards, transparency and human accountability.
2. Core Principle
AI should assist human decision-making. It should not become an unchecked autonomous authority.
Aivry systems are designed to augment people and business workflows rather than remove lawful human accountability.
The level of human oversight required should increase as the potential impact of an AI-supported decision increases.
3. Nature of AI Systems
Aivry systems may use technologies including:
- Large language models
- Speech recognition
- Speech synthesis
- Transcription systems
- Retrieval-augmented generation
- Knowledge retrieval
- Workflow automation
- Prompt orchestration
- Classification systems
- Recommendation systems
- APIs
- Third-party AI models
- Customer-provided knowledge sources
Depending on the deployment, underlying technology may be provided by third-party providers such as OpenAI, Anthropic, Google, ElevenLabs, Sarvam AI and other technology, cloud, telecommunications or infrastructure providers.
Not every provider is used in every Aivry product or deployment.
4. AI Limitations
Artificial intelligence systems are probabilistic.
Outputs may occasionally include:
- Inaccurate information
- Incomplete information
- Hallucinated content
- Outdated information
- Inconsistent responses
- Irrelevant responses
- Biased suggestions
- Incorrect classifications
- Misinterpretations
- Contextual mistakes
- Formatting errors
- Incorrect workflow decisions
AI is not infallible.
Customers must independently validate AI outputs where reliance on those outputs could materially affect individuals, operations, finances, healthcare, employment, compliance or legal obligations.
5. Human Oversight
Customers must maintain appropriate human oversight where AI-supported decisions may materially affect people or organizations.
Examples include:
- Hiring decisions
- Candidate rejection
- Healthcare communication
- Medical escalation
- Customer complaints
- Financial approvals
- Contract interpretation
- Regulatory communication
- Legal notices
- Compliance actions
- Account suspension
- Employee termination
- High-value transactions
- Safety-critical operational decisions
AI outputs should not be blindly relied upon in high-impact situations.
6. No Autonomous Professional Advice
Unless expressly designed, validated and legally permitted for a particular purpose, Aivry systems are not professional advisors.
Aivry does not guarantee:
- Legal advice
- Medical advice
- Accounting advice
- Financial advice
- Tax advice
- Regulatory advice
- Employment-law advice
- Compliance certification
Where an AI response touches a regulated or professional subject, users should obtain appropriate qualified human review.
7. Transparency
Where reasonably appropriate, people interacting with Aivry-powered systems should be able to understand that they are interacting with an automated or AI-assisted system.
Customers must not intentionally configure Aivry systems to deceptively impersonate a real individual.
AI-generated voices, chatbots or automated agents must not be used to falsely represent themselves as a specific real person without lawful authority.
Transparency requirements may vary depending on:
- Communication channel
- Applicable law
- Industry
- Type of interaction
- User expectation
- Customer configuration
8. Recruitment AI Principles – Halo
Halo is designed to assist recruitment teams, not replace them.
Halo may assist with:
- Resume parsing
- Candidate analysis
- Candidate organization
- Skills matching
- AI-assisted scoring
- Recruitment communication
- Workflow assistance
Halo does not:
- Make final employment decisions
- Autonomously hire candidates
- Autonomously reject candidates
- Replace human recruitment accountability
- Guarantee that candidate rankings are free from bias
Candidate scoring, ranking or recommendations generated through Halo are advisory only.
Customers remain responsible for:
- Reviewing candidates
- Validating scoring criteria
- Checking for inappropriate bias
- Maintaining meaningful human oversight
- Equal-opportunity compliance
- Employment-law compliance
- Anti-discrimination compliance
- Candidate privacy
Customers should also refer to the Halo AI Recruitment Fair Use Addendum.
9. Bias and Fairness
AI systems may reflect or amplify bias.
Bias may arise from:
- Training data
- Third-party models
- Customer-provided data
- User prompts
- Uploaded documents
- Incomplete context
- Incorrect assumptions
- Selection criteria
- Historical business practices
Customers using AI in people-impacting workflows must actively consider fairness.
This is particularly important for:
- Recruitment
- Candidate screening
- Customer segmentation
- Customer escalation
- Healthcare communication
- Eligibility decisions
- Prioritization systems
Aivry does not guarantee completely bias-free output.
Human review must remain available wherever automated output could materially affect an individual.
10. Milo Healthcare AI Principles
Milo is designed primarily as an AI patient experience and administrative communication system.
Milo may assist healthcare providers with:
- Appointment booking
- Appointment confirmations
- Calendar coordination
- Patient reminders
- Pre-appointment instructions
- Follow-up communication
- Patient enquiries
- CRM updates
- Administrative workflow coordination
Milo should operate within clearly defined boundaries established by the healthcare provider.
Milo must not independently:
- Diagnose a medical condition
- Prescribe medication
- Recommend treatment as a substitute for a clinician
- Alter prescribed treatment
- Make autonomous clinical decisions
- Decide whether emergency medical treatment is required without an appropriate escalation mechanism
- Represent itself as a qualified doctor or healthcare professional
- Override instructions provided by qualified medical personnel
Human Escalation
Milo implementations should provide appropriate escalation to human staff where:
- The patient requests human assistance
- The AI cannot confidently respond
- The request falls outside approved knowledge
- Symptoms or medical concerns are raised
- The conversation indicates an emergency or potentially urgent condition
- The AI detects conflicting or unclear information
- Clinical judgment may be required
Milo should not be treated as an emergency-response system.
Source Control
Medical or procedural information communicated through Milo should be based on material approved by the healthcare customer.
Healthcare providers remain responsible for ensuring that:
- Instructions are accurate
- Information remains current
- Medical content is appropriately approved
- Patient communication complies with applicable requirements
11. Atithi Hospitality AI Principles
Atithi is designed to assist hotels, resorts and hospitality businesses with guest communication and operational coordination.
Atithi may assist with:
- Guest enquiries
- Concierge communication
- Reservation-related information
- Food and beverage requests
- Housekeeping requests
- Maintenance requests
- Service routing
- Request tracking
- Internal coordination
- Guest follow-ups
Operational Accuracy
Atithi may depend on information supplied by:
- Hotel staff
- Property-management systems
- Booking systems
- CRM systems
- POS systems
- Internal databases
- Customer-provided documents
AI-generated responses are only as reliable as the underlying information available.
Atithi should not independently guarantee:
- Room availability
- Upgrade availability
- Pricing
- Reservation confirmation
- Service completion
- Food availability
- Maintenance completion
- Special requests
unless the relevant information has been confirmed through the appropriate connected system.
Human Escalation
Guest conversations should be escalated where appropriate, including:
- Complaints
- Safety issues
- Medical emergencies
- Security concerns
- Payment disputes
- Sensitive guest matters
- Requests outside approved operating parameters
Aivry AI assists communication. The hospitality customer remains responsible for delivering the physical guest experience.
12. Messaging and Voice Safety
Where Aivry systems interact through WhatsApp, voice calls, SMS, email or other communication channels, customers remain responsible for:
- Lawful opt-ins
- Consent
- Required disclosures
- Anti-spam compliance
- Telecom compliance
- Call-recording requirements
- Platform policies
- Applicable messaging rules
Aivry systems must not be used for:
- Spam
- Harassment
- Phishing
- Deceptive outreach
- Unlawful impersonation
- Unauthorized bulk messaging
- Unlawful automated calling
AI-generated voices should not be used to deceptively imitate identifiable individuals without appropriate authorization.
13. Automation Safety
Automation creates operational risk irrespective of which Aivry product initiates it.
Possible risks include:
- Incorrect triggers
- Unintended messages
- Duplicate messages
- Incorrect CRM updates
- Data corruption
- Incorrect routing
- Failed API actions
- Repeated actions
- Escalation loops
- Workflow loops
- Incorrect booking actions
- Erroneous status changes
- Third-party integration failures
Customers should validate workflows before production deployment.
High-impact automations should include appropriate safeguards such as:
- Human approval
- Confirmation steps
- Access restrictions
- Input validation
- Logging
- Rollback procedures
- Escalation paths
- Rate limits
- Error handling
Automation should never remove human oversight where a failed action could materially affect a person or business.
14. Knowledge Source Integrity
Aivry systems may generate responses based on customer-supplied information.
This may include:
- Websites
- PDFs
- Knowledge bases
- FAQs
- Policies
- Product catalogues
- Medical instructions
- Hospitality information
- CRM records
- Pricing
- Internal documents
- Operational procedures
If these materials are inaccurate, incomplete, contradictory or outdated, AI responses may inherit those problems.
Customers remain responsible for the quality and accuracy of information supplied to their Aivry systems.
Aivry may provide tools or processes to update knowledge sources, but the customer remains responsible for approving authoritative business information.
15. Data Minimisation
AI systems should process only information reasonably necessary for their intended function.
Customers should avoid providing unnecessary:
- Personal information
- Sensitive personal information
- Medical information
- Candidate information
- Financial information
- Authentication credentials
- Confidential business information
where such information is not needed for the relevant workflow.
Access to sensitive data should be appropriately restricted.
16. Sensitive Data
Where Aivry processes sensitive information, additional safeguards may be appropriate.
This is particularly relevant for:
- Patient information
- Healthcare-related information
- Candidate information
- Financial information
- Identity documents
- Authentication data
- Confidential customer records
Customers remain responsible for determining whether their proposed use is lawful and appropriate.
Aivry may impose additional technical, contractual or operational requirements for sensitive deployments.
17. Third-Party Model Dependency
Aivry may rely on third-party AI infrastructure.
Capabilities and behaviour may change due to:
- Model updates
- Provider policy changes
- Rate limits
- Downtime
- Pricing changes
- API modifications
- Safety-system updates
- Model behaviour changes
- Feature deprecation
Aivry cannot guarantee that a third-party AI model will behave identically over time.
Where appropriate, Aivry may modify prompts, orchestration, model selection or infrastructure to maintain service quality and safety.
18. Security Expectations
Customers must not attempt to:
- Extract confidential system prompts
- Circumvent safeguards
- Manipulate models for malicious purposes
- Gain unauthorized access
- Abuse credentials
- Probe protected infrastructure without authorization
- Introduce malicious instructions or payloads
- Exploit vulnerabilities
- Perform destructive automation
- Access data belonging to another customer
Security abuse may result in immediate restriction or suspension.
19. Prohibited AI Uses
Aivry AI systems must not be used for:
Fraud
Including:
- Phishing
- Scams
- Fraudulent communication
- Identity impersonation
- Deceptive financial requests
Harmful Conduct
Including:
- Harassment
- Threats
- Abuse
- Malicious targeted communication
- Deliberate harmful misinformation
Illegal Activity
Including:
- Unlawful surveillance
- Rights violations
- Regulatory evasion
- Unauthorized access
- Illegal data collection
Discrimination
Including:
- Unlawful employment discrimination
- Automated exclusion based on protected characteristics
- Prohibited profiling
Unsafe Professional Decisions
Including autonomous high-impact:
- Medical decisions
- Legal decisions
- Employment decisions
- Financial decisions
- Regulatory decisions
without appropriate qualified oversight.
Manipulation
Aivry systems must not be intentionally configured to materially deceive vulnerable individuals or exploit their lack of understanding that they are interacting with an automated system.
20. High-Impact Decisions
AI should not be the sole determinant of decisions that could significantly affect an individual’s:
- Employment
- Healthcare
- Legal rights
- Access to essential services
- Financial status
- Safety
Where Aivry technology assists such workflows, appropriate human review must remain available.
21. Human Handover
Aivry encourages human handover where:
- The AI cannot confidently answer
- A user explicitly requests a person
- The interaction becomes sensitive
- A dispute develops
- The matter involves professional judgment
- Safety may be affected
- The AI detects uncertainty or conflicting information
- The situation exceeds the approved scope of the AI system
Customers are responsible for establishing practical escalation processes appropriate to their deployment.
22. Monitoring and Quality Review
Aivry and its customers may use appropriate monitoring mechanisms to identify:
- Repeated hallucinations
- Incorrect responses
- Workflow failures
- Unsafe AI behaviour
- Abuse
- Security incidents
- Inappropriate bias
- Failed escalations
- Misconfigured prompts
- Knowledge-source problems
Where appropriate and permitted by applicable agreements and law, system logs and interactions may be reviewed for troubleshooting, quality assurance, security and safety purposes.
23. Incident Reporting
Customers should notify Aivry where they identify material issues including:
- Harmful AI behaviour
- Serious hallucinations
- Patient-safety concerns
- Problematic recruitment behaviour
- Unsafe guest interactions
- Major automation failures
- Security incidents
- Unauthorized access
- Significant data exposure
- Repeated incorrect AI responses
Reports may be submitted to:
grievance@getaivry.com
24. Continuous Improvement
Responsible AI is an ongoing process.
Aivry may improve:
- System prompts
- AI orchestration
- Guardrails
- Human-handover logic
- Retrieval systems
- Workflows
- Monitoring
- Abuse detection
- Model selection
- Security controls
- Integration safeguards
to improve safety, accuracy, reliability and usefulness.
Changes may be implemented as technology, regulation and customer requirements evolve.
25. Customer Responsibility
Aivry provides the technology, but customers control many aspects of deployment.
Customers are responsible for:
- Selecting appropriate use cases
- Providing lawful data
- Providing accurate knowledge
- Configuring appropriate permissions
- Maintaining human oversight
- Reviewing material AI decisions
- Complying with industry regulations
- Obtaining necessary consent
- Monitoring deployed systems
- Training relevant staff
- Establishing escalation procedures
AI should not be deployed in a high-risk workflow without considering the consequences of incorrect output.
26. Limitation of Responsible AI Controls
Responsible-AI safeguards can reduce risk.
They cannot eliminate all AI risk.
No currently available generative AI system can guarantee:
- Perfect accuracy
- Zero hallucination
- Complete neutrality
- Zero bias
- Continuous availability
- Perfect interpretation of intent
Customers should design their operational processes accordingly.
27. Relationship With Other Aivry Policies
This Responsible AI Policy should be read together with:
- Privacy Policy
- Terms & Conditions
- Acceptable Use Policy
- Cookie Policy
- Halo AI Recruitment Fair Use Addendum
- Applicable Data Processing Agreements
- Applicable customer agreements and Statements of Work
Where a separately executed customer agreement establishes additional AI safeguards, those requirements will also apply.
28. Updates to This Policy
Aivry may update this Responsible AI Policy to reflect:
- New products
- New AI capabilities
- Regulatory developments
- Changes in technology
- New safety practices
- Changes to deployment models
- Lessons from operating AI systems
The current version will be published on Aivry’s website with the applicable last-updated date.
29. Contact
Questions, concerns or reports relating to responsible AI may be directed to:
Adzura Media (OPC) Private Limited
Operating under the brand name Aivry
Flat No. 402, Pandit Rao Nilayam
East Marredpally
Secunderabad, Telangana – 500026
India
Email: grievance@getaivry.com
