Responsible AI and Ethical Use Statement

2. Our Responsible AI Principles

Our work is guided by seven principles:

• Accountability and governance
• Human oversight and appropriate use
• Transparency and disclosure
• Fairness, accessibility, and non-discrimination
• Privacy and responsible data use
• Security, resilience, and product safety
• Accuracy, testing, and ongoing review

These principles apply throughout the system lifecycle, from planning and data selection through deployment, operation, modification, and retirement.

3. Accountability and Governance

Web Inventix AI accepts responsibility for the work and decisions within our control.

For each AI project, we aim to define:

• The business problem and intended outcome
• The system’s users and affected parties
• The role performed by AI
• The data required
• The actions the system may take
• The decisions that require human approval
• Known limitations and foreseeable failure scenarios
• Client, vendor, and Web Inventix AI responsibilities
• Escalation, incident, and human handoff procedures

We apply a risk-based review before deployment. Systems that may affect a person’s rights, employment, finances, healthcare, safety, access to services, or other significant interests require deeper assessment and stronger controls.

4. Shared Responsibility

Responsible AI requires cooperation between Web Inventix AI, the client, technology providers, data providers, system users, and other parties involved in the project.

Web Inventix AI may rely on third-party models, hosting providers, communication services, databases, APIs, and software platforms. These providers control parts of their infrastructure, model behaviour, service availability, data handling, and product changes.

Clients remain responsible for matters within their control, including:

• Providing lawful and accurate data
• Obtaining required notices, permissions, and consents
• Assigning qualified human reviewers
• Using the system only for its approved purpose
• Following operating and security instructions
• Reviewing material AI-generated outputs before acting on them
• Reporting errors, misuse, security events, or unexpected results
• Meeting industry-specific and professional obligations

Project agreements may assign more detailed responsibilities.

5. Human Oversight and Appropriate Use

AI should support human judgement rather than remove accountability from people or organizations.

We design human review and approval controls based on the risk of the use case. Controls may include:

• Human approval before an action is completed
• Escalation to a staff member
• Manual review of low-confidence results
• Restrictions on the systems and records an AI agent may access
• Approval requirements for financial, contractual, or customer-facing actions
• Logs showing significant system activity
• Methods to correct, reverse, or override system actions

We do not recommend using AI as the sole decision-maker for high-impact decisions involving:

• Employment or worker evaluation
• Credit, lending, insurance, or financial eligibility
• Healthcare diagnosis or treatment
• Legal rights or legal advice
• Housing
• Education admissions or assessment
• Government benefits or public services
• Personal safety
• Other decisions that may significantly affect an individual

6. Transparency and Disclosure

People should receive clear information when they interact with an AI system or when AI materially affects a service or decision.

Depending on the use case, we may recommend disclosure of:

• The fact that the person is interacting with AI
• The system’s purpose
• The general role AI performs
• The categories of information being collected
• How information may be used
• Known limitations
• The availability of human assistance
• How a person can question or challenge a result

We do not intentionally present AI systems as human beings in a deceptive manner.

AI voice agents, chatbots, and other public-facing agents should identify themselves as automated or AI-assisted when appropriate to the context, risk, and applicable requirements.

Transparency does not require the disclosure of source code, confidential business information, security-sensitive details, or third-party proprietary information.

7. Fairness, Accessibility, and Non-Discrimination

AI systems can reproduce errors, exclusions, and biases found in data, system design, human decisions, or third-party models.

For relevant projects, we may assess:

• The suitability and representation of data
• Differences in performance across user groups
• Accessibility barriers
• The effect of incorrect classifications or recommendations
• The risk of unfair treatment
• The availability of human review or alternative service channels

We do not knowingly design or deploy AI for unlawful discrimination, unfair profiling, or treatment based on protected personal characteristics.

Systems serving the public should provide a practical method for people to obtain human assistance when automated interaction is unsuitable or inaccessible.

When a material fairness problem is identified, we may adjust the data, prompts, rules, thresholds, workflow, user interface, or system scope.

8. Privacy and Responsible Data Use

We apply privacy-by-design practices appropriate to the system, information, and level of risk.

Our practices may include:

• Collecting only the information needed for the approved purpose
• Defining permitted uses before data is processed
• Limiting data retention
• Applying role-based access controls
• Separating client environments and records
• Using de-identified, anonymized, or synthetic data when practical
• Reviewing third-party data handling and retention practices
• Documenting cross-border processing where relevant
• Providing methods for access, correction, or deletion where required
• Completing privacy or data assessments for higher-risk projects

Information entered into third-party AI services may be processed under those providers’ terms, privacy policies, settings, and data-retention practices. We assess these factors when selecting technology for a project.

Sensitive information, including health, financial, biometric, employment, identity, and children’s information, requires additional review and restrictions.

9. Data and Intellectual Property

Data, documents, images, software, and other content used in an AI system must come from authorized sources.

We aim to:

• Confirm the client’s right to provide and use project data
• Respect confidentiality and intellectual property obligations
• Follow relevant licences and usage restrictions
• Avoid using unlawfully obtained personal or proprietary information
• Document significant external knowledge sources when appropriate
• Protect client-owned prompts, workflows, business rules, and materials according to the project agreement

AI-generated content may contain errors, similarities to existing content, or uncertain ownership status. Clients should review important content before publishing, licensing, selling, or relying on it.

10. Security, Resilience, and Product Safety

We apply security controls based on the system’s risk, architecture, data, integrations, and operating environment.

Controls may include:

• Authentication and access management
• Least-privilege permissions
• Encryption in transit and at rest where supported and appropriate
• Secure credential and secret management
• Environment separation
• Activity logging
• Input and output restrictions
• Rate limits and spending limits
• Backup and recovery procedures
• Software dependency management
• Security testing proportionate to the project
• Incident response and containment procedures

We limit agent permissions to approved tools, records, actions, and workflows. Higher-risk actions may require human approval, secondary validation, transaction limits, or other safeguards.

Security reviews may consider AI-specific risks such as prompt injection, unauthorized instructions, data leakage, excessive permissions, unsafe tool use, malicious files, unreliable outputs, and attempts to bypass system controls.

11. Accuracy, Reliability, and Testing

AI output may be incorrect, incomplete, outdated, inconsistent, or inappropriate for the intended purpose.

Before deployment, testing may include:

• Normal operating scenarios
• Incomplete or unclear inputs
• Incorrect or conflicting information
• Attempts to bypass instructions
• Low-confidence conditions
• Escalation and human handoff
• Privacy and security failures
• System outages or unavailable integrations
• Harmful, discriminatory, or misleading outputs

For higher-risk systems, testing may also include representative datasets, documented acceptance criteria, structured evaluations, red-team testing, independent review, or controlled pilot deployment.

AI-generated information should be verified before it is used for legal, medical, financial, employment, safety, contractual, or other significant decisions.

We do not guarantee that an AI system will produce error-free results.

12. Monitoring, Changes, and Incident Response

AI models, third-party platforms, data sources, integrations, and user behaviour can change over time.

Depending on the engagement, ongoing controls may include:

• Performance monitoring
• Error and escalation tracking
• Cost and usage monitoring
• Review of model or vendor changes
• User feedback
• Quality sampling
• Security alerts
• Version records
• Rollback procedures
• Periodic reassessment of the approved use case

Material incidents should be investigated and documented. Corrective action may include restricting system access, disabling a feature, correcting data, changing prompts or rules, notifying affected parties, contacting a provider, or suspending the system.

Incident notification obligations depend on the nature of the event, applicable law, and the project agreement.

13. Uses We Will Not Knowingly Support

Web Inventix AI will not knowingly design, configure, or deploy AI systems for:

• Illegal activity
• Fraud, identity theft, or deceptive impersonation
• Unlawful surveillance or tracking
• Unlawful discrimination or prohibited profiling
• Harassment, exploitation, or targeted abuse
• Deliberate creation or distribution of materially false information intended to cause harm
• Unauthorized access to systems or information
• Re-identification of protected de-identified information without lawful authority
• Fully autonomous high-impact decisions without appropriate human control
• Uses that materially conflict with human rights, privacy rights, safety, or applicable law

14. Continuous Improvement

Responsible AI practices will change as technology, research, laws, industry standards, and business uses develop.

We review this statement periodically and may update it when:

• Our services change
• Material legal or regulatory requirements change
• New risks or technical controls emerge
• Client or user feedback identifies a gap
• A significant incident or system failure provides new information

This statement is informed by recognized AI risk-management, privacy, security, and human-rights principles. It does not represent a certification under a particular framework unless Web Inventix AI expressly confirms that certification in writing.

15. Questions and Concerns

Questions, complaints, or concerns about Web Inventix AI’s responsible AI practices may be sent to:

We will review credible reports and determine the appropriate response based on the system, project, risk, contractual responsibilities, and applicable requirements.

16. Status of This Statement

This statement describes Web Inventix AI’s general responsible AI principles.

It does not replace:

• A client agreement
• Statement of work
• Data-processing agreement
• Privacy policy
• Security requirements
• Acceptable-use policy
• Industry-specific obligations
• Applicable laws or professional advice

Specific project commitments are governed by the applicable written agreement.