Definitions

"PDL" means PsyData Labs L.L.C.
"Model" means any AI model, model weights, embeddings, fine-tuned model, inference system, AI agent, evaluation model, or model artifact owned or developed by PDL and subject to this license.
"Model Weights" means the numerical parameters constituting the trained state of the Model, which are treated as proprietary trade secrets.
"Inference" means the execution of the Model to generate outputs from inputs.
"Fine-Tuning" means additional training of the Model on new data to adapt its outputs.
"Knowledge Distillation" means the process of training a smaller or separate model to replicate the behavior or outputs of the PDL Model.
"Model Cloning" means any technique to reproduce, replicate, or reconstruct Model Weights or architecture.
"Commercial Inference Deployment" means running the Model to serve requests in a commercial product, service, or API.
"Responsible AI" means ethical, safe, and accountable development and deployment of AI systems consistent with applicable law, professional standards, and PDL policy.
"Prohibited Use" means any use of the Model as enumerated in Section 12.

License Grant

Subject to the terms herein, PsyData Labs L.L.C. grants Licensee a limited, non-exclusive, non-transferable license to use the Model solely for the specific purpose and within the specific constraints set forth in Schedule A. This license authorizes only: (a) local or private Inference use within the authorized scope; and (b) fine-tuning, if expressly authorized in Schedule A. All other rights are expressly reserved by PsyData Labs L.L.C.

Restrictions

The following are strictly PROHIBITED under this Model License:

Redistribution of Model Weights in any form, medium, or format;
Extraction, Knowledge Distillation, Model Cloning, or training of competing models using PDL Model outputs, weights, activations, or architecture;
Reverse engineering of Model architecture from outputs, embeddings, or API responses;
Unauthorized Commercial Inference Deployment without a separate written Commercial Inference License;
Fine-tuning the Model without express written authorization in Schedule A;
Sharing, publishing, or otherwise making Model Weights available to third parties;
Using the Model for any Prohibited Use as enumerated in Section 17 (AI-Specific Terms). 13. Attribution Requirements
All deployments, publications, and communications referencing or utilizing the Model must include the Standard PDL Attribution Block. Model cards and technical reports involving the Model must acknowledge PsyData Labs L.L.C. as the model developer.

Ownership & Intellectual Property

PDL Model Weights are the exclusive proprietary trade secrets of PsyData Labs L.L.C. No ownership interest in the Model, Model Weights, or underlying architecture is transferred by this license. Licensee-created fine-tuned model artifacts, if fine-tuning is authorized, may be jointly governed by this Agreement and a supplemental agreement as defined in Schedule A.

Confidentiality

Model Weights, architecture specifications, evaluation benchmarks, training methodologies, and inference configurations constitute Confidential Information and trade secrets. Licensee shall implement technical and organizational measures to prevent unauthorized access to or disclosure of Model Weights. Confidentiality obligations survive termination indefinitely for trade secret protections.

Data Protection

Any personal data, PHI, or PII processed through Model Inference must be handled in compliance with applicable data protection law. Licensee must ensure that Model Inference pipelines do not store or log sensitive user data beyond operational necessity and must comply with applicable retention and deletion obligations.

AI-Specific Terms

This section sets forth mandatory Responsible AI obligations applicable to all Licensees of PDL Models:

17.1 Responsible AI Obligations
Licensee must implement ongoing bias monitoring and fairness evaluation procedures for any Model deployment;
Safety evaluation and red-teaming must be conducted before any deployment involving high-stakes decisions;
Output verification procedures must be established to prevent propagation of harmful or erroneous outputs;
Incident reporting: Licensee must report significant AI failures, unexpected outputs, safety incidents, or harmful outputs to PDL within seventy-two (72) hours of discovery.
17.2 Output Disclaimer Requirements
Licensee must not present Model outputs as authoritative clinical, diagnostic, psychological, legal, or financial conclusions without qualified professional review. Applicable disclosures that Model outputs are AI-generated and require expert verification must be prominently displayed in all clinical, legal, and financial applications.

17.3 Model Card Requirements
Commercial deployments must maintain a current model card documenting: intended use, out-of-scope use, training data overview, evaluation results, known limitations, and bias considerations, consistent with established model documentation standards.

17.4 Safety Red-Teaming
Prior to production deployment, Licensee must conduct or commission adversarial safety evaluation to identify failure modes, prompt injection vulnerabilities, and harmful output pathways. Results must be documented and made available to PDL upon request.

17.5 Prohibited Uses
Use of the Model for any of the following is strictly PROHIBITED:

Mass surveillance or individual surveillance without informed consent;
Psychological manipulation of individuals or groups without informed consent;
Discriminatory profiling based on protected characteristics;
Clinical diagnosis or treatment decisions without qualified human clinical oversight;
Military weapons development or autonomous lethal systems;
Any application violating applicable law;
Deceptive AI personas designed to deceive users into believing they are interacting with humans. 18. Commercial Terms
Commercial Inference Deployment requires a separate written Commercial Inference License executed by an authorized officer of PDL. Commercial terms, inference volume limits, and fees are defined in Schedule A. Unauthorized commercial inference constitutes a material breach.

Redistribution Terms

Redistribution of Model Weights in any form is PROHIBITED. Redistribution of fine-tuned model artifacts requires separate written authorization from PDL and is subject to the same use restrictions as this Model License.

Modification Terms

Fine-tuning and modification of the Model are permitted only if expressly authorized in Schedule A. All modifications remain subject to PDL Intellectual Property Rights. Licensee must not use fine-tuning to circumvent Responsible AI obligations or Prohibited Use restrictions.

Termination

This Agreement terminates immediately upon: (a) redistribution of Model Weights; (b) Knowledge Distillation or Model Cloning; (c) violation of Prohibited Use provisions; or (d) breach of Confidentiality obligations. For other material breaches, PDL shall provide written notice and a thirty (30) day cure period. Upon termination, Licensee must immediately cease all Inference, delete or destroy all copies of Model Weights, and certify destruction in writing to PDL.

Warranty Disclaimer

THE MODEL IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND. PDL MAKES NO WARRANTY THAT THE MODEL'S OUTPUTS ARE ACCURATE, SAFE, UNBIASED, OR FIT FOR ANY PARTICULAR PURPOSE, INCLUDING CLINICAL, DIAGNOSTIC, LEGAL, OR FINANCIAL APPLICATIONS. LICENSEE ASSUMES ALL RISK ASSOCIATED WITH DEPLOYMENT AND USE OF THE MODEL.

Limitation of Liability

PDL'S AGGREGATE LIABILITY UNDER THIS MODEL LICENSE SHALL NOT EXCEED THE GREATER OF LICENSE FEES PAID OR $100.00 USD. PDL IS NOT LIABLE FOR ANY HARM RESULTING FROM MODEL OUTPUTS, BIAS, ERRORS, OR FAILURES, INCLUDING CLINICAL HARM, FINANCIAL LOSS, OR DISCRIMINATION.

Indemnification

Licensee shall indemnify PDL from all claims arising from Model deployment, Prohibited Use, clinical or legal misapplication of outputs, failure to implement Responsible AI safeguards, or breach of this Agreement.

Governing Law

This Agreement shall be governed by the laws of Madison County, New York State, United States.

Contact Information

PsyData Labs L.L.C. | legal@psydata.org | https://www.psydata.net

Project-Specific Schedule — Schedule A

Field Value
Project Name [MODEL NAME]
Version [MODEL VERSION]
Authorized Use [INFERENCE / RESEARCH / FINE-TUNING — specify]
Commercial Threshold [API CALLS / REQUESTS PER MONTH]
Royalty Rate [ROYALTY RATE OR "N/A"]
Revenue Share % [REVENUE SHARE % OR "N/A"]
Audit Rights Reserved by PsyData Labs L.L.C. upon reasonable written notice
Expiration Date [EXPIRATION DATE]