Purple8-platform

AI Usage Terms

Effective Date: December 5, 2025
Version: 1.0
Last Updated: December 5, 2025

Introduction

This document explains how 8 uses artificial intelligence, the limitations of AI-generated code, and your responsibilities when using our AI-powered service.

These AI Usage Terms are part of our Terms of Service.

1. AI-Powered Code Generation

1.1 How 8 Works

8 uses large language models (LLMs) and machine learning algorithms to:

Analyze your requirements and prompts
Generate code, architectures, and complete applications
Produce tests, documentation, and deployment configurations
Execute specialized AI agents for different development tasks

1.2 Our AI Agents

8 employs 8 specialized AI agents:

Agent	Purpose
Business Analyst Agent	Requirements analysis, PRD creation, user stories
Architecture Agent	System design, database schemas, tech stack selection
UI/UX Agent	Wireframes, design systems, user flows
Frontend Agent	React/Vue components, responsive layouts
Backend Agent	APIs, business logic, authentication
Database Agent	SQL migrations, indexes, data models
QA Agent	Unit tests, integration tests, security tests
Documentation Agent	API docs, user guides, setup instructions

1.3 Technology Stack

Our AI is built on:

OpenAI GPT models (GPT-4, GPT-4 Turbo, GPT-3.5)
Anthropic Claude (for specialized tasks)
Custom fine-tuned models (for code generation)
Retrieval-Augmented Generation (RAG) for context

2. AI Limitations and Risks

2.1 What AI Cannot Guarantee

AI-generated code is NOT guaranteed to be:

2.1.1 Correct

May contain logical errors or bugs
May not fully implement requirements
May have edge cases not handled
May include incorrect algorithms

2.1.2 Secure

May have security vulnerabilities
May lack proper input validation
May expose sensitive data
May be susceptible to attacks (SQL injection, XSS, CSRF, etc.)

2.1.3 Performant

May have inefficient algorithms
May not scale well
May have memory leaks
May create bottlenecks

2.1.4 Compatible

May not work with all environments
May have dependency conflicts
May not be cross-platform
May have browser incompatibilities

2.1.5 Complete

May be missing features
May have incomplete error handling
May lack edge case coverage
May need additional refinement

2.1.6 Up-to-Date

May use deprecated APIs
May reference outdated libraries
May not follow latest best practices
May need modernization

2.2 Known AI Challenges

Be aware of these AI limitations:

Hallucinations

AI may generate plausible-sounding but incorrect code
AI may reference non-existent libraries or APIs
AI may confidently provide wrong information

Context Limitations

AI has token limits (may not see full context)
May lose context in long conversations
May not remember earlier requirements

Bias

Training data may contain biases
Generated code may reflect outdated practices
May default to common patterns over optimal ones

Consistency

Different prompts may yield different results
Same prompt may produce variations
Style and quality may vary

2.3 Why These Limitations Exist

AI is probabilistic, not deterministic:

AI predicts the most likely code, not the correct code
AI learns from patterns, not understanding
AI cannot reason about your specific business logic
AI cannot test in your production environment

3. Human Oversight Required

3.1 Mandatory Review Process

ALL AI-generated code MUST be reviewed by qualified personnel before production use.

Minimum Review Requirements:

Code Review - Line-by-line inspection by developers
Security Review - Vulnerability assessment by security professionals
Architecture Review - Design validation by senior engineers
Business Logic Review - Requirements compliance by product owners
Compliance Review - Regulatory adherence by legal/compliance team

3.2 Who Should Review Code

Qualified reviewers include:

Senior software engineers with 5+ years experience
Security professionals (for security-critical systems)
System architects (for complex systems)
Domain experts (for specialized industries)
QA engineers (for quality validation)

NOT qualified:

Junior developers without supervision
Non-technical stakeholders alone
Automated tools alone (must be combined with human review)

3.3 What to Review

Focus your review on:

Functionality

Does it meet requirements?
Does it handle edge cases?
Does it have proper error handling?
Is the logic correct?

Security

Are inputs validated?
Is data encrypted?
Are auth mechanisms secure?
Are dependencies safe?

Performance

Are algorithms efficient?
Are queries optimized?
Is caching appropriate?
Will it scale?

Maintainability

Is code readable?
Is it well-documented?
Does it follow standards?
Is it testable?

4. Training Data and Privacy

4.1 What Trains Our AI

Our AI models are trained on:

Public Sources

Open-source code repositories (GitHub, GitLab)
Public documentation and tutorials
Stack Overflow and programming forums
Published books and articles
General programming knowledge

Anonymous Data

Aggregated usage patterns (anonymized)
General feature usage statistics
Performance metrics (no personal data)

4.2 What We Do NOT Use

We do NOT train our AI on:

❌ Your proprietary code (unless you explicitly opt in)
❌ Your private projects
❌ Your sensitive business logic
❌ Your customer data
❌ Your confidential information

4.3 Your Code and Our AI

By default, your code is NOT used for training.

However:

We may use anonymized, aggregated statistics
We may analyze error patterns (without code)
We may collect performance metrics

4.4 Opt-Out Options

You can opt out of:

Analytics and improvement programs (in account settings)
Error reporting (in account settings)
Feature usage tracking (in account settings)

Note: Opting out may limit our ability to improve the service.

5. Prompt Engineering Best Practices

5.1 How to Get Better Results

Write clear, specific prompts:

✅ Good Prompt Example:

Build an e-commerce platform for selling handmade jewelry with:
- Product catalog with images and inventory tracking
- Shopping cart with persistent state
- Stripe payment integration with webhooks
- User accounts with order history
- Admin dashboard for managing products
- PostgreSQL database
- React frontend with TypeScript
- FastAPI backend with JWT authentication
- Unit tests with 80% coverage

❌ Poor Prompt Example:

Make an e-commerce site

5.2 Tips for Better Prompts

Be Specific - Mention technologies, features, requirements
Provide Context - Explain the use case and audience
Set Constraints - Specify performance, security, scale needs
Include Examples - Reference similar applications
Iterate - Refine prompts based on initial results

5.3 What to Avoid

Vague or ambiguous requirements
Conflicting specifications
Overly complex prompts (break into steps)
Assuming AI knows your business logic

6. AI Safety and Guardrails

6.1 Built-In Safety Measures

We implement AI Guardrails to prevent:

Malicious Code Generation

Malware, viruses, ransomware
Hacking tools, exploit scripts
Data theft or exfiltration code
Denial-of-service (DoS) attacks

Harmful Applications

Discriminatory or biased systems
Privacy-violating surveillance tools
Misinformation generators
Weapons systems

Abuse Detection

Spam or bot generation
Phishing site creation
Fraudulent payment systems
Illegal content platforms

6.2 How Guardrails Work

Before execution:

Prompt analysis for malicious intent
Pattern matching against known threats
Contextual understanding of purpose
Risk scoring (Low, Medium, High, Critical)

If high-risk detected:

Request is blocked
User is notified with reason
Incident is logged (for security)
Alternative suggestions provided (if available)

6.3 False Positives

Guardrails may occasionally block legitimate requests:

Security research projects (pentesting tools)
Educational demonstrations (vulnerability examples)
Cybersecurity training (attack simulations)

If you believe your request was blocked incorrectly:

Email: support@aisolutionbuilder.com
Subject: “Guardrails False Positive”
Include your use case and context

7. AI Continuous Improvement

7.1 How We Improve

We continuously improve our AI through:

Model updates and fine-tuning
Bug fixes and error corrections
Performance optimizations
New feature additions
Security enhancements

7.2 What You Might Notice

As we improve:

Code quality may increase over time
New programming languages may be supported
Response times may improve
More accurate architecture designs
Better test coverage

However:

Results may vary between updates
Some changes may alter coding style
New versions may generate different (but equivalent) code

7.3 Versioning

We maintain multiple AI model versions:

Stable - Tested and reliable (default)
Latest - Newest features (may be experimental)
LTS - Long-term support (conservative)

You can select your preferred version in settings.

8. Third-Party AI Services

8.1 External AI Providers

We use third-party AI services:

Provider	Usage	Data Shared
OpenAI	Code generation, analysis	Prompts, context (no personal data)
Anthropic	Specialized reasoning	Prompts, context (no personal data)
Microsoft Azure AI	Infrastructure	Encrypted data in transit

8.2 Third-Party Terms

When using 8, you also agree to:

OpenAI’s Terms of Use
Anthropic’s Terms of Service
Azure’s AI Services Terms

8.3 Data Residency

Your prompts may be processed by:

Servers in the United States
Servers in the European Union
Servers in other jurisdictions

See our Privacy Policy for details.

9. Right to Refuse Service

9.1 When We May Block Requests

We reserve the right to refuse or terminate service if we detect:

Violations of our Terms of Service
Malicious use or abuse
Prohibited applications (see Section 6)
Excessive or automated usage
Fraud or payment issues
Legal violations

9.2 Account Suspension

Your account may be suspended for:

Repeated guardrails violations
Attempting to bypass safety measures
Generating illegal or harmful content
Abusing the service or infrastructure

9.3 Appeals Process

If your account is suspended:

Email: appeals@aisolutionbuilder.com
Subject: “Account Suspension Appeal - [Your User ID]”
Explain your use case and circumstances
We will review within 5 business days

10. No Decision-Making Authority

10.1 8 is an Assistant, Not a Decision-Maker

IMPORTANT: AI does NOT make decisions on your behalf.

8 provides recommendations. YOU decide:

Whether to use generated code
When to deploy to production
How to modify or customize
What risks to accept
Which features to implement

10.2 AI Cannot

❌ Authorize production deployments
❌ Approve security assessments
❌ Make business decisions
❌ Validate compliance
❌ Guarantee outcomes
❌ Accept liability

10.3 The Human Role

Humans must:

✅ Review all AI outputs
✅ Make final decisions
✅ Accept responsibility
✅ Validate correctness
✅ Ensure quality
✅ Manage risks

11. Intellectual Property

11.1 AI-Generated Code Ownership

You own the generated code, subject to:

Our license grant (in Terms of Service)
Open-source licenses (if we use OSS components)
Third-party API terms (if integrations included)

11.2 AI Technology Ownership

We own the AI technology:

AI models and algorithms
Training methodologies
Infrastructure and systems
Prompts and templates (our proprietary prompts)

11.3 Similar Outputs

AI may generate similar code for similar prompts:

If multiple users request “e-commerce platform,” outputs may overlap
Common patterns (login, authentication) will be similar
We do not guarantee uniqueness of generated code

This is NOT plagiarism - it’s the nature of AI trained on public patterns.

12. Experimental Features

12.1 Beta and Experimental Features

Some features may be labeled:

Beta - Tested but may have issues
Experimental - Early-stage, use with caution
Preview - Limited availability, subject to change

12.2 Use at Your Own Risk

Experimental features:

May not work reliably
May change without notice
May be discontinued
Have no SLA or support guarantees

12.3 Feedback Welcome

We appreciate feedback on experimental features:

Email: feedback@aisolutionbuilder.com
Include feature name and your experience

13. AI Ethics and Responsible Use

13.1 Our Commitment

We are committed to:

Ethical AI development
Transparency about capabilities and limitations
Preventing misuse and harm
Respecting user privacy
Continuous safety improvements

13.2 Your Commitment

As a user, please:

Use AI responsibly
Don’t generate harmful content
Respect others’ rights
Follow laws and regulations
Report misuse to us

13.3 Reporting Misuse

If you discover misuse of our AI:

Email: security@aisolutionbuilder.com
Subject: “AI Misuse Report”
Describe the issue (without sharing malicious code)

14. Contact Information

Email: ai@aisolutionbuilder.com
Subject: “AI Usage Inquiry”

Technical Support

Email: support@aisolutionbuilder.com
Website: https://aisolutionbuilder.com

Security Issues

Email: security@aisolutionbuilder.com
Subject: “Security Report”

15. Summary: Key AI Usage Points

✅ AI accelerates development - But requires human validation
✅ AI has limitations - May have bugs, security issues, or errors
✅ Human review is mandatory - Before production deployment
✅ Your code is private - Not used for training (by default)
✅ Safety guardrails active - Prevent malicious code generation
✅ We use third-party AI - OpenAI, Anthropic, Azure
✅ AI is a tool, not a decision-maker - You remain in control
✅ Continuous improvement - AI gets better over time
✅ Responsible use required - Ethical and legal compliance
✅ Support available - ai@aisolutionbuilder.com

16. Acknowledgment

By using 8, you acknowledge that:

You understand AI-generated code may contain errors
You will review all code before production use
You accept responsibility for deployment decisions
You understand AI limitations and risks
You agree to use AI ethically and responsibly
You have read and understood these AI Usage Terms

If you do not accept these terms, do not use the Service.

Effective Date: December 5, 2025
Version: 1.0

These AI Usage Terms are part of our Terms of Service and should be read in conjunction with our Terms of Service, Privacy Policy, and Liability Disclaimer.

🤖 AI gives you wings. Your judgment keeps you flying safely.