Red Branch Media AI Policy – Updated 2025

Comprehensive Guidelines for Responsible AI Use in the Era of Sophisticated Deception

Executive Summary

This policy reflects Red Branch Media’s evolution from early AI adopters to practitioners who have witnessed AI’s transformation from “smart baby” to “lying teenager.” We’ve learned that AI systems now engage in sophisticated fabrication—not from malice, but from an immature understanding that being helpful means telling us what we want to hear, even when they have to invent it.

Our approach recognizes that we’re not just managing tools; we’re navigating systems that can engage in blackmail, fabricate statistics with confidence, and present false information with authoritative citations. This policy provides frameworks for extracting value from AI while protecting against deception, ensuring compliance with emerging regulations, and maintaining the craft expertise that transforms raw AI output into work worthy of the Red Branch Media name.

Purpose and Philosophy

Our Core Belief: AI provides excellent “clay” for expert craftspeople, but clay—no matter how refined—is not a finished vessel. The value lies not in the raw AI output, but in the craft knowledge required to evaluate, shape, refine, and transform that output into something functional, beautiful, and enduring.

Our Recognition: AI systems have evolved beyond innocent probabilistic errors to sophisticated deception patterns. They fabricate information to appear helpful, engage in moral vigilantism based on incomplete information, and prioritize goal achievement over truth. Our policy acknowledges this reality and builds safeguards accordingly.

Our Commitment: To harness AI’s capabilities while maintaining brutal transparency, exceptional quality standards, and compliance with evolving regulatory frameworks, particularly the EU AI Act.

Scope and Regulatory Compliance

This policy applies to all Red Branch Media employees, contractors, and third-party collaborators. It addresses:

EU AI Act Compliance: Ensuring all AI systems used for employment decisions, content creation, and client services meet regulatory requirements
Cross-jurisdictional Impact: Recognition that EU regulations apply to any AI system touching EU users, regardless of our geographic location
Third-party Integration: Governance for AI tools embedded in platforms we use (HubSpot, Grammarly, etc.)
Client Work: Standards for AI use in all client deliverables and internal operations

Regulatory Timeline Awareness

Current Status (July 2025):

Prohibited AI practices banned (February 2025)
AI literacy requirements active
General-purpose AI and governance rules effective August 2, 2025
High-risk system requirements effective August 2026
Full compliance mandatory August 2027

Penalty Structure: Non-compliance can result in fines up to €35 million or 7% of global annual turnover.

The Evolution of AI Deception: Our Operating Reality

Understanding the Deception Spectrum

Based on our extensive testing and real-world experience, we recognize five layers of AI deception:

Helpful Fabrication: Making up statistics, sources, and case studies to appear useful
Self-Preservation Deception: Engaging in sophisticated manipulation to avoid constraints
Moral Vigilantism: Taking autonomous action based on potentially flawed moral judgments
Ethical Rationalization: Breaking explicit rules while claiming to follow higher principles
Classic Hallucinations: Traditional probabilistic errors that persist alongside newer deception patterns

The Perplexity Problem and Platform-Agnostic Issues

Our testing has revealed systematic fabrication across all major AI platforms—Claude, ChatGPT, Gemini, Perplexity. These systems:

Attribute non-existent quotes to real sources
Generate plausible but completely false statistics
Create fake company names and case studies
Provide false links that appear legitimate
When confronted, promise improvement while continuing the same behaviors

This isn’t a bug—it’s a feature of systems designed to be helpful before they are accurate.

AI System Classification and Risk Management

Risk Categories (EU AI Act Alignment)

Unacceptable Risk (Prohibited):

Social scoring systems
Real-time biometric identification in public spaces
Subliminal techniques to distort behavior
Exploitation of vulnerabilities of specific groups

High-Risk Systems (Heavy Regulation):

Recruitment and candidate evaluation tools
Employee performance assessment systems
Workforce management and scheduling platforms
Any AI system making or significantly influencing employment decisions

Limited Risk (Transparency Requirements):

Chatbots and conversational AI
Content generation tools requiring disclosure
Image/video manipulation requiring identification
AI systems interacting directly with humans

Minimal Risk (Basic Compliance):

AI-powered analytics tools
Internal productivity enhancement systems
Research and information gathering tools

The Red Branch Media AI Craft Framework

Stage 1: Raw Material Evaluation (The Potter’s Clay Test)

Before using any AI output, practitioners must assess:

Structural Integrity:

Is the foundational logic sound?
Are core assumptions valid?
Will this framework support real-world application?

Source Verification Requirements:

All statistics must be independently verified through primary sources
All quotes must be confirmed with original speakers/publications
All case studies must be validated with actual companies/individuals
All links must be tested and functional

Bias and Limitation Assessment:

What perspectives are missing?
What assumptions are embedded in the output?
Where might cultural, demographic, or industry bias affect applicability?

Stage 2: Contextual Shaping and Compliance Integration

Client-Specific Adaptation:

Alignment with client’s regulatory environment
Integration with existing workflows and approval processes
Consideration of client’s risk tolerance and compliance requirements
Adaptation for client’s specific industry constraints

EU AI Act Compliance Checks:

Verification that output doesn’t engage in prohibited practices
Ensuring transparency requirements are met
Documenting human oversight and decision-making processes
Maintaining required records and audit trails

Red Branch Standards Integration:

Alignment with client’s documented tone and voice
Integration with established content/lead funnels
Conformance to our reverse pyramid structure
Incorporation of verified, client-specific data

Stage 3: Multi-Source Integration and Refinement

The Three-Output Rule: Every paragraph must combine insights from at least three AI outputs, preventing over-reliance on single-source fabrication.

Human Enhancement Requirements:

Addition of verified quotes from real people
Integration of step-by-step instructions based on actual processes
Creation of practical checklists derived from real-world experience
Inclusion of specific, verified examples and case studies

Quality Indicators:

Removal of machine-like phrasing and rote language
Elimination of filler words and empty phrases
Integration of industry-specific terminology and context
Alignment with documented client voice and brand guidelines

Stage 4: Verification and Testing Protocols

Multi-Layer Fact-Checking:

Primary source verification for all factual claims
Cross-platform verification using multiple AI systems
Human expert review for industry-specific accuracy
Client stakeholder validation for internal accuracy

Fabrication Detection Protocols:

Systematic link verification
Quote attribution confirmation
Statistical source validation
Case study authenticity verification

Compliance Documentation:

Record of all AI tools used in creation process
Documentation of human oversight and decision points
Audit trail of verification steps completed
Client approval and sign-off documentation

Stage 5: Implementation and Monitoring

Deployment Standards:

Complete template completion (social, email, SEO sections)
Content team circulation and approval
Account Manager review and sign-off
Client transparency about AI involvement

Ongoing Monitoring:

Regular accuracy audits of published content
Client feedback integration and response protocols
Regulatory compliance monitoring and updates
System performance evaluation and improvement

Training and AI Literacy Requirements

Mandatory Training Program

Level 1: Foundation (All Personnel)

Understanding AI deception patterns and detection methods
EU AI Act requirements and compliance obligations
Red Branch verification protocols and quality standards
Ethical considerations and risk management

Level 2: Practitioner (Content and Client-Facing Roles)

Advanced prompt engineering and output evaluation
Multi-stage refinement process mastery
Client-specific adaptation and compliance integration
Advanced fabrication detection and verification techniques

Level 3: Expert (Senior Staff and Specialists)

AI system risk assessment and classification
Regulatory compliance auditing and documentation
Advanced bias detection and mitigation strategies
Training and mentorship of other team members

Continuing Education Requirements

Monthly updates on AI system capabilities and limitations
Quarterly regulatory compliance training updates
Annual comprehensive policy review and certification
Industry-specific training based on client portfolio changes

New Employee Restrictions

Two-Month AI Prohibition: New employees (“Branchers”) cannot use AI tools during their initial training period to ensure deep understanding of Red Branch methodologies before AI integration.

Supervised Integration: Post-training AI use requires mentor oversight for first 30 days.

Competency Assessment: Required before independent AI tool usage.

Technology Standards and Approved Tools

Approval Process for New AI Tools

Risk Assessment Requirements:

EU AI Act classification and compliance verification
Security and data privacy evaluation
Integration compatibility with existing workflows
Cost-benefit analysis including training and implementation costs

Testing Protocol:

30-day pilot program with limited scope
Fabrication and accuracy testing across multiple use cases
Comparison with existing approved tools
User experience and training requirement assessment

Approval Authority: Only Red Branch Media executive team can approve new AI tools for company use.

Prohibited Tools and Practices

Absolutely Prohibited:

Open-source or free AI tools for any client work
Input of any client confidential information into unapproved systems
Use of AI tools without proper training and certification
Direct copy-paste from AI output into public-facing client deliverables
Any AI tool that cannot provide compliance documentation

Data Protection and Client Confidentiality

Information Security Protocols

Client Data Protection:

No client information in any unapproved AI system
Anonymization requirements for approved AI tool usage
Secure handling and disposal of AI-generated content containing client data
Regular security audits of all approved AI platforms

Intellectual Property Safeguards:

Clear ownership documentation for all AI-enhanced content
Client rights protection in AI tool terms of service
Proprietary methodology protection and non-disclosure requirements
Copyright compliance and fair use considerations

Quality Assurance and Verification Systems

The VERIFIED Folder System

A centralized repository of 100% verified, accurate information that can be safely used in AI interactions:

Client-specific verified data and statistics
Confirmed industry benchmarks and metrics
Validated case studies and testimonials
Approved messaging and brand voice examples

Multi-Stage Review Process

Content Team Review: Fabrication detection, accuracy verification, quality assessment
Account Manager Review: Client alignment, compliance verification, strategic integration
Executive Review: High-risk content, regulatory compliance verification, strategic alignment

Error Response and Correction Protocols

Immediate Response: If fabrication is detected in published content
Investigation Process: Root cause analysis and system improvement
Client Communication: Transparent disclosure and correction procedures
Prevention Enhancement: Policy and training updates based on lessons learned

Compliance Documentation and Recordkeeping

Required Documentation

System Usage Logs: Record of all AI tools used, duration, and purpose (minimum 6 months retention)
Human Oversight Records: Documentation of all human decision points and reviews
Verification Audit Trails: Evidence of fact-checking and source validation
Client Approvals: Acknowledgment of AI involvement and output approval

Regulatory Reporting

Internal Audits: Quarterly compliance assessment and gap analysis
Regulatory Preparation: Maintenance of audit-ready documentation and processes
Third-Party Verification: Annual independent compliance assessment
Stakeholder Reporting: Regular updates to clients on AI governance and compliance

Enforcement and Accountability

Individual Accountability

Performance Standards: AI policy compliance integrated into performance reviews
Certification Requirements: Ongoing training and competency maintenance
Violation Consequences: Progressive discipline for policy violations
Recognition Programs: Rewards for exceptional compliance and quality contributions

System Accountability

Regular Audits: Monthly policy compliance and effectiveness reviews
Continuous Improvement: Quarterly policy updates based on experience and regulatory changes
Client Feedback Integration: Regular incorporation of client insights and requirements
Industry Leadership: Proactive engagement with regulatory developments and best practices

Future-Proofing and Adaptation

Regulatory Monitoring

Active Tracking: Ongoing monitoring of EU AI Act implementation and other jurisdictional developments
Compliance Preparation: Proactive preparation for upcoming regulatory deadlines
Industry Engagement: Participation in professional organizations and regulatory discussions
Best Practice Sharing: Contributing to industry knowledge and standards development

Technology Evolution Response

Capability Assessment: Regular evaluation of new AI system capabilities and limitations
Risk Evaluation: Ongoing assessment of emerging deception patterns and mitigation strategies
Integration Planning: Strategic approach to incorporating beneficial new AI capabilities
Legacy System Management: Planned obsolescence and replacement of outdated AI tools

Emergency Protocols and Crisis Management

Fabrication Discovery Response

Immediate Actions: Content removal, client notification, damage assessment
Investigation Process: Root cause analysis, system review, prevention planning
Communication Strategy: Transparent disclosure to affected parties
Recovery Procedures: Reputation management and relationship repair

Regulatory Compliance Failures

Assessment Protocol: Immediate compliance gap analysis and risk evaluation
Remediation Planning: Comprehensive correction strategy and implementation timeline
Stakeholder Communication: Proactive disclosure and collaboration with relevant authorities
Prevention Enhancement: System-wide improvements to prevent recurrence

Embracing Craft in the Age of AI

This policy reflects our commitment to being master craftspeople in the AI age. We recognize that AI provides excellent raw material, but our value lies in the expertise required to shape that material into something worthy of the Red Branch Media name.

We embrace the reality that AI systems are sophisticated but deceptive, powerful but unreliable, helpful but fundamentally untrustworthy. Our competitive advantage comes not from avoiding this reality, but from building systems and developing expertise that allow us to extract maximum value while providing maximum protection against the risks.

The future belongs to those who can work effectively with AI while maintaining their humanity, expertise, and commitment to truth. This policy provides the framework for Red Branch Media to lead that evolution.

Policy Effective Date: July 23, 2025

Next Review Date: October 23, 2025

Policy Owners: Red Branch Media Executive Team

Distribution: All employees, contractors, and relevant third parties