AI Transparency Disclosure

Honest AiB Holdings, Inc. believes AI should show its work. This disclosure describes how artificial intelligence is used across our products, what it can and cannot do, and how we ensure transparency, accuracy, and accountability in every AI-generated output.

This disclosure is aligned with the transparency requirements of the EU AI Act (Regulation 2024/1689) and reflects our commitment to responsible AI deployment.

AI Systems We Use

Our products use third-party large language models (LLMs) and vision models as processing engines. Currently, these include:

• Google Gemini: Used for natural language parsing, document extraction, event recognition, and data classification across our product suite.
• Anthropic Claude: Used for complex analysis, verification, and counter-argument generation.
• OpenAI GPT models: Tested and validated as part of our multi-LLM verification framework.

Our proprietary AiBcheck verification protocol has been tested and validated across 8 major LLMs.

What AI Does in Our Products

cAIlendar: AI parses natural language text messages and photographs to extract calendar event data (event name, date, time, participants). AI determines a confidence score for each extraction. Low-confidence extractions require human confirmation before being saved.

hGrAiL Legal Billing Audit: AI reads legal billing invoices, extracts line items, matches them against LEDES standards and engagement parameters, and flags anomalies. Every AI-identified anomaly is presented with the source data and confidence level for human review.

TaxhAck Ai: AI analyzes tax return data to identify potential errors, missed deductions, and optimization opportunities. All findings include source references and are presented as advisory — not as tax advice.

GAAP App: AI assists in classifying transactions, preparing financial statements, and checking GAAP compliance. All outputs require professional CPA review before filing.

verAIcity / AiBcheck Protocol: AI generates responses that include numbered source citations, confidence scoring, identified knowledge gaps, counter-arguments, and self-verification. The system is designed to surface uncertainty rather than hide it.

What AI Cannot Do

Our Verification Framework

Every AI output in our products passes through the AiBcheck verification architecture:

• Source Citation: Every claim is tagged with a numbered source — logic, external data, or user input. Uncited claims are flagged.
• Confidence Scoring: Multi-dimensional scoring evaluates response reliability. When confidence drops below acceptable thresholds, the system triggers additional review or responds with "I don't know."
• Gap Identification: Every output identifies what the AI does not know or cannot verify about a given topic.
• Counter-Argument Validation: The AI challenges its own conclusions with opposing evidence. Both sides are presented to the user.
• Self-Verification: Every output concludes with an internal consistency check. Missing verification invalidates the output.
• AiBGUARD™ Footer: A mandatory verification footer appears on every output, reminding the user to verify independently.

Audit Trail (hGrAiL™)

Every AI interaction is logged with:

• The original input (what the user provided)
• The AI extraction or analysis (what the AI interpreted)
• The confidence level (how certain the AI was)
• The human verification status (confirmed, modified, or rejected)
• A timestamp and source attribution

For products using Google Sheets as the data layer, this audit trail is visible and editable by the user at all times. For enterprise products, audit logs are available upon request.

Human Oversight

Our products are designed with human-in-the-loop verification at every critical decision point. AI proposes; humans confirm. No AI output in our system is treated as final without human review for decisions affecting legal, financial, tax, or compliance outcomes.

Data and Training

We do not use your data to train AI models. Your inputs are processed by third-party AI providers under their respective data processing agreements, which prohibit use of API inputs for model training. See our Privacy Policy for full data handling details.

EU AI Act Alignment

Honest AiB Holdings, Inc. is committed to compliance with the EU AI Act (Regulation 2024/1689). Our products are designed to meet the Act's requirements for:

• Transparency (Article 50): Users are always informed that AI is processing their data, and how
• Human Oversight (Article 14): Critical decisions require human confirmation
• Technical Documentation (Articles 11–12): This disclosure and our audit trail architecture provide required documentation
• Record-Keeping (Article 12): hGrAiL™ audit trails maintain comprehensive logs of AI system operation
• Accuracy and Robustness (Article 15): Confidence scoring and counter-argument validation identify and communicate uncertainty

Reporting Concerns

If you believe an AI output from any Honest AiB product was incorrect, harmful, or insufficiently transparent, please contact us immediately. We take every report seriously and will investigate the specific output using our audit trail.

Honest AiB Holdings, Inc.
Email: jon@aibcheck.com
Web: aibcheck.com