# W3rocks

# llms.txt — LLM disclosure for w3rocks.com
# Version: 1.0
# Last-Updated: 2025-10-17
# Format: simple key: value pairs + repeated blocks for models. Comments start with '#'.

Site: w3rocks.com
Contact: privacy@w3rocks.com
Maintainer: W3Rocks Product Team
Purpose: Transparency about large language models (LLMs) used on this site and their roles.

####################
# MODEL BLOCK (1)
####################
Model-Name: OpenAI GPT-4o
Provider: OpenAI
Provider-URL: https://openai.com
Model-Version: GPT-4o (commercial API)
Deployment: API (HTTPS)
Primary-Use-Cases:
  - Conversational assistant on marketing.w3rocks.com
  - Generating marketing copy, email templates, and chat replies
  - Lead enrichment suggestions
Training-Data-Description:
  - Trained by provider on a mixture of licensed data, human-created data, and publicly available text (see provider model card)
Safety-Notes:
  - Rate-limited; content filters applied via provider moderation API and additional server-side checks
  - Sensitive-data redaction applied before storage (PII minimization)
Model-Card: https://platform.openai.com/docs/model-cards/gpt-4o
Confidence-Level: high

####################
# MODEL BLOCK (2)
####################
Model-Name: Llama 2 (13B)
Provider: Meta / third-party host
Provider-URL: https://ai.meta.com/llama
Model-Version: Llama-2-13b
Deployment: Self-hosted / third-party inference (as configured)
Primary-Use-Cases:
  - On-prem or private inference for lead scoring experiments
  - Local prompt experimentation for agents and rule-based augmentation
Training-Data-Description:
  - Publicly released model from Meta; trained on a mixture of public and licensed data (see model card)
Safety-Notes:
  - Access limited to internal services; content filtered before display
Model-Card: https://ai.meta.com/llama/model-card
Confidence-Level: medium

####################
# MODEL BLOCK (3)
####################
Model-Name: Open-Source Embedding Models
Provider: Various (OpenAI, Hugging Face, Cohere, etc.)
Model-Version: v* (depends on job)
Deployment: API or self-hosted
Primary-Use-Cases:
  - Semantic search over site content and lead data
  - Similarity matching for contact and prospect enrichment
Training-Data-Description:
  - Provider-specific; mostly public and licensed datasets
Safety-Notes:
  - Embedding vectors stored encrypted; PII removed where possible
Model-Card: see provider documentation
Confidence-Level: medium

####################
# DATA SOURCES & USAGE
####################
Input-Data-Types:
  - Public web pages, user-submitted content (forms, chat), CRM records submitted by users, customer-provided files (only with explicit consent)
Persistent-Storage:
  - Transcripts and generated outputs may be stored for up to 90 days for debugging and quality; user request can remove stored records (see Data Rights below)
PII-handling:
  - We attempt to minimize PII in prompts; fields marked 'sensitive' are redacted prior to external API calls when feasible
Derivative-Use:
  - Outputs may be used to improve system prompts and internal analytics (aggregated, not tied to user identity)

####################
# SAFETY & MITIGATIONS
####################
Content-Filtering:
  - Provider moderation APIs are invoked for potentially disallowed content (violence, hate, sexual content, illicit instructions)
  - Additional server-side heuristics block or flag disallowed outputs
Human-in-the-Loop:
  - High-risk actions (legal, medical, financial advice) flagged for human review before sending actionable recommendations
Transparency:
  - Generated content is labelled "AI-generated" where output might be used in customer-facing communications (emails, proposals)
Vulnerability-Disclosure:
  - Security issues can be reported to security@w3rocks.com

####################
# USER RIGHTS & CONTROLS
####################
Opt-Out:
  - To opt out of non-essential personalization and data retention, contact privacy@w3rocks.com with subject "LLM OPT-OUT"
Data-Deletion:
  - Users can request deletion of their stored chat transcripts and associated data. Email privacy@w3rocks.com with "DELETE MY LLM DATA"
Correction:
  - If an AI-produced record contains incorrect personal data, request correction via privacy@w3rocks.com
Explainability:
  - For model decisions (e.g., why a lead was scored as high), we provide a human-readable explanation on request within 7 business days

####################
# MODEL CARDS & LINKS
####################
OpenAI Model Card: https://platform.openai.com/docs/model-cards
Llama Model Card: https://ai.meta.com/llama/model-card
Provider-Docs: see model-specific URLs above

####################
# LEGAL & LICENSE
####################
License: Use of models shown is subject to providers' Terms of Service and this site's Privacy Policy.
Privacy-Policy: https://w3rocks.com/privacy
Terms-of-Service: https://w3rocks.com/terms

####################
# NOTES
####################
- This file is a transparency statement only and does not include confidential implementation details (API keys, private endpoints).
- To update this file, modify and redeploy to https://w3rocks.com/.well-known/llms.txt and set Last-Updated to the new date.
- If you need a machine-parseable JSON or YAML variant, request "llms.json" or "llms.yaml".

# End of file