01. Development and testing
I develop these tools as a licensed attorney with a background in computer science and software development. Testing includes automated checks, source and citation review, and trial runs using public information or synthetic test data.
Testing recorded as of October 28, 2026 includes:
- Claim Charting Tool: 52 automated tests passed. Checks also cover citations and compatibility with saved charts. Testing of the full chart-revision process is not yet complete.
- Interactive Art Unit Predictor: 74 Python tests passed, along with tests of the recommendation rules. These cover selected goals, combined amendments, source checks, and error handling. The accuracy of the art unit estimates and reported probabilities has not yet been established.
- IDF Enhancement Tool: 37 unit and regression tests and 11 database integration tests passed. Sign-in, reviewer permissions, uploads, and AI processing were also checked with test data.
The Open-Source Legal Workflow Builder is planned and has not yet been built. These tests do not amount to an independent security audit or guarantee that the tools’ legal analysis is correct.
02. No-training and Zero Data Retention
No-training means the model provider does not use your inputs or the model’s outputs to train its models. It does not mean the provider keeps no data.
Zero Data Retention (ZDR) means the provider processes the request without keeping its input and output afterward, under the terms of its ZDR service. ZDR must be enabled for the particular service being used.
The data still goes to the provider for processing. The app may also save information, and providers may keep technical records such as usage data. Exceptions may apply for legal requirements and certain safety checks. See OpenRouter’s explanation of ZDR and its notice about screening and legal retention requirements.
03. Data settings by tool
These are the settings in use as of October 28, 2026:
| Tool | How model data is handled |
|---|---|
| Claim Charting Tool | Inputs and outputs are not used to train models. ZDR is enabled. |
| IDF Enhancement Tool | Inputs and outputs are not used to train models. ZDR is enabled. |
| Interactive Art Unit Predictor | Inputs and outputs are not used to train models. ZDR is not enabled. This version is for public information or synthetic test data only. |
ZDR covers model inputs and outputs. It does not cover data saved by the app or handled by search and file-processing services.
We have checked the tools’ settings, but have not audited the model providers. Their terms still apply.
04. Storage and access
The tools may save files, drafts, questions and answers, charts, sources, account information, and other records needed to operate. ZDR for model calls does not delete those records. Hosting, database, backup, email, search, and file-processing services have their own data policies.
The IDF tool uses email-code sign-in, separate client workspaces, assigned reviewers, and its own database. The Claim Charting Tool includes access controls for invited reviewers. The Art Unit Predictor is currently a local development preview; it does not yet offer public accounts.
Model API keys are stored on the server and are not included in browser code. Checks cover how credentials and access permissions are handled. No online service can guarantee against unauthorized access, outages, data loss, or configuration errors.
05. What you may submit
Use public information or synthetic test data that you are authorized to share. Do not submit confidential client information, privileged documents, or other sensitive material. No-training and ZDR settings do not create privilege or an attorney–client relationship.
Verify the outputs before using them. If your organization requires particular storage locations, retention periods, contracts, or security standards, do not assume these tools meet them. Read the Disclaimer & conditions of use before using a tool.