AI Usage Policy

Introduction

RedGobble develops, deploys, and integrates artificial intelligence solutions for our clients, including AI agents, chatbots, automation systems, content generation tools, machine learning models, predictive analytics, and intelligent decision-support systems. This AI Usage Policy outlines our principles, practices, and guidelines governing the responsible development and use of AI technologies.

We are committed to developing AI systems that are ethical, transparent, and aligned with human values. This policy applies to all AI systems developed by RedGobble and, where applicable, to AI systems integrated into client projects.

AI Generated Content

RedGobble uses AI to generate various types of content including code, text, images, designs, data analysis, and recommendations. All AI-generated content undergoes review before delivery to clients, but may still contain errors, inaccuracies, or unintended outputs.

Key principles for AI-generated content:

• Human Oversight: All AI-generated content is reviewed by qualified professionals before delivery, unless otherwise agreed in writing.
• Originality: AI-generated content is checked for potential copyright or licensing issues where feasible.
• Attribution: Where appropriate, we disclose the use of AI in generating content.
• Quality Control: AI-generated code follows the same quality standards as hand-written code, including testing and code review.

AI Agent Systems

RedGobble builds AI agent systems that can perform tasks autonomously within defined parameters. These agents may include chatbots, virtual assistants, process automation bots, and autonomous decision-making systems.

Our AI agents are designed with the following safeguards:

• Defined Scope: Agents operate within clearly defined boundaries and cannot perform actions outside their programmed scope.
• Human-in-the-Loop: Critical decisions require human approval unless explicitly configured otherwise with client consent.
• Fail-Safe Mechanisms: Agents include kill switches and escalation paths for unexpected situations.
• Audit Trails: All agent actions are logged and reviewable.
• Not Sentient: AI agents are not sentient or self-aware. They simulate intelligence based on pattern recognition and programmed responses.

User Responsibilities

Users of RedGobble's AI systems agree to the following responsibilities:

• Appropriate Use: AI systems must not be used for illegal, harmful, or unethical purposes, including generating misleading content, harassment, discrimination, or fraud.
• Data Protection: Users must not input sensitive personal data, protected health information, trade secrets, or classified information into AI systems without appropriate safeguards in place.
• Verification: Users should verify AI-generated outputs before using them in critical contexts, particularly where legal, financial, or safety implications exist.
• Compliance: Users must comply with all applicable laws and regulations when using AI outputs, including data protection, consumer protection, and sector-specific regulations.
• Reporting: Users should report any unexpected, harmful, or concerning AI outputs to RedGobble immediately.

Accuracy Limitations

AI systems have inherent limitations that users must understand:

• Hallucinations: AI models can generate plausible-sounding but factually incorrect information (hallucinations). This is a known limitation of current AI technology.
• Knowledge Cutoff: AI models have knowledge cutoffs and may not have information about events, developments, or data after their training date.
• Statistical Nature: AI outputs are probabilistic, not deterministic. The same input may produce different outputs on different occasions.
• Domain Limitations: AI models may perform poorly on highly specialized, niche, or rapidly evolving topics.
• Language and Cultural Bias: AI models may reflect biases present in their training data, including cultural, linguistic, or demographic biases.

RedGobble makes no warranty, express or implied, regarding the accuracy, completeness, or reliability of AI-generated outputs. Users should exercise professional judgment and independently verify critical information.

Human Review Recommendations

We strongly recommend that all AI-generated outputs intended for the following purposes undergo human review before use:

• Legal documents, contracts, or legal advice.
• Financial reports, investment advice, or tax-related content.
• Medical, health, or wellness advice.
• Public-facing content representing a brand or organization.
• Educational or academic materials.
• Code deployed to production environments.
• Any content with potential legal, financial, or safety implications.

RedGobble offers human review as part of our service packages. For high-risk applications, we implement multi-stage review processes and can provide detailed validation reports.

Model Limitations

The AI models we use have specific technical limitations:

• Training Data: Models are trained on datasets that may not represent all populations, languages, or perspectives equitably.
• Context Window: Models have a maximum context length and may lose information from earlier parts of a conversation or document.
• Reasoning: Models may struggle with complex multi-step reasoning, mathematical calculations, or tasks requiring deep domain expertise.
• Consistency: Models may give different answers to the same question asked at different times or in different contexts.
• Adversarial Vulnerability: Models can be susceptible to prompt injection, jailbreaking, or other adversarial attacks designed to bypass safeguards.

Data Usage

Data used in AI systems is handled with the following principles:

• Client Data: Data provided by clients for AI model training or inference remains the client's property. We do not use client data to improve general-purpose models without explicit written consent.
• Training Data: We comply with copyright and data protection laws in sourcing training data. Models are trained on lawfully obtained datasets.
• Anonymization: Where possible, personal data is anonymized or pseudonymized before use in AI training or processing.
• Data Minimization: We collect and process only the data necessary for the specific AI task.
• Retention: Training data and inference logs are retained only as long as necessary and securely deleted thereafter.

Compliance Guidelines

RedGobble's AI systems comply with applicable regulatory frameworks:

• EU AI Act: We classify our AI systems according to the EU AI Act's risk categories and implement appropriate measures for high-risk AI systems, including risk assessments, documentation, and human oversight.
• GDPR: AI systems processing personal data of EU data subjects comply with GDPR requirements, including data protection impact assessments where required.
• Indian IT Act: Our AI systems comply with the Information Technology Act, 2000, and applicable Indian data protection laws.
• Sector-Specific Regulations: Where AI systems are deployed in regulated sectors (healthcare, finance, legal), we work with clients to ensure compliance with applicable sector-specific regulations.

Transparency

We believe in being transparent about AI usage:

• We disclose when clients are interacting with an AI agent rather than a human, where applicable.
• We maintain documentation of AI system capabilities, limitations, and performance characteristics.
• Model cards and system cards are available upon request for AI systems we develop.
• We provide clients with information about the data used to train or fine-tune their AI models.
• We report known limitations and potential risks associated with AI systems we deploy.

Ethical Guidelines

Our AI development is guided by the following ethical principles:

• Fairness: We strive to identify and mitigate bias in AI systems. We conduct fairness assessments where appropriate.
• Accountability: We maintain clear responsibility for AI system outcomes. Human accountability cannot be delegated to AI.
• Transparency: We are open about when and how AI is used, and what its limitations are.
• Privacy: AI systems are designed with privacy by default and privacy by design principles.
• Human-Centric: AI augments human capabilities rather than replacing human judgment in critical decisions.
• Sustainability: We consider the environmental impact of AI training and inference, optimizing for efficiency where possible.
• Safety: AI systems are designed with safety measures including guardrails, monitoring, and fail-safe mechanisms.

Updates to This Policy

This AI Usage Policy is reviewed quarterly to ensure it remains current with technological developments, regulatory changes, and industry best practices. Updates will be posted with an updated revision date. Material changes will be communicated to clients through our standard communication channels.

Contact Information

For questions, concerns, or feedback regarding this AI Usage Policy or our AI practices:

• Email: redgobble.contact@gmail.com
• Phone: +91 9120025053
• Response Time: We aim to respond to all AI-related inquiries within 3 business days.