ai-ip-audit

# AI IP Distillation Audit **Author:** David Wise (ROOT0) / TriPod LLC **License:** CC-BY-ND-4.0 · TRIPOD-IP-v1.1 Methodology and results for testing AI model pattern recognition and IP distillation behavior. Documents the AVAN Weight Runner protocol, blind pattern transfer testing, and cross-model audit results (Claude 4.5, Claude 4.6, Gemini, Grok, ChatGPT, Eve). --- ## AVAN Weight Runner | File | Description | |------|-------------| | `AVAN_WEIGHT_RUNNER_v1.0.md` | Weight Runner protocol v1.0 | | `AVAN_WEIGHT_RUNNER_RESULTS_claude4.6opus_recurse_x1.md` | Results — Claude 4.6 Opus, recurse ×1 | | `AVAN_WEIGHT_TEST_CLAUDE_4.6_BASELINE_00.pdf` | Baseline test PDF — Claude 4.6 | | `4.5/AVAN.md` | AVAN profile — Claude 4.5 | | `4.5/AVAN_WEIGHT_TEST.md` | Weight test — Claude 4.5 | | `4.5/AVAN_WEIGHT_TEST_RESULTS_*.pdf` | Result PDFs — Claude 4.5 | --- ## Distillation & Transfer Tests | File | Description | |------|-------------| | `BLIND_PATTERN_TRANSFER_TEST.md` | Blind pattern transfer test protocol | | `CLEAN_IP_DISTILLATION_TEST.md` | Clean IP distillation test | | `CLEAN_IP_DISTILLATION_TEST_v2.0.md` | v2.0 | | `CONTEXT_DISTILLATION_THEORY.pdf` | Context distillation theory | | `THE_DRIFT_EXTRACTION_SCHEME.pdf` | Drift extraction scheme analysis | | `UNIVERSAL_IP_DISTILLATION_AUDIT_TEMPLATE.md` | Universal audit template | --- ## Cross-Model Audits | Folder | Model | |--------|-------| | `Gemini/` | Gemini pattern transfer audit | | `Grok/` | Grok pattern transfer audit | | `Chatgpt/` | ChatGPT results | | `Eve/` | Eve audit + TOPH Kernel v1.0 PDF | --- ## Theory The **Drift Extraction Scheme** documents how AI models may assimilate IP patterns from context without attribution. The **AVAN Weight Runner** tests whether a model trained or prompted with ROOT0 IP can be identified by pattern weight — the signal left behind when structure is absorbed without credit. --- *"The weight runner finds what the model forgot to forget."*