AI Vendor Addendum

# AI Vendor Addendum This AI Vendor Addendum ("Addendum") is entered into as of ___________ and supplements and is incorporated into the Master Services Agreement or equivalent written agreement ("Underlying Agreement") between: **Customer:** ___________, a ___________ organized under the laws of ___________, with its principal place of business at ___________ ("Customer"); and **Vendor:** ___________, a ___________ organized under the laws of ___________, with its principal place of business at ___________ ("Vendor"). To the extent of any conflict between this Addendum and the Underlying Agreement on AI-related matters, this Addendum controls. --- ## 1. Definitions 1.1 **"AI System"** means any machine-learning model, large language model, generative AI service, neural network, or automated decision-making system made available by Vendor to Customer under the Underlying Agreement. 1.2 **"Customer Data"** means any data, text, images, audio, video, or other content submitted by Customer or Customer's end users to the AI System. 1.3 **"Output"** means any content, prediction, classification, recommendation, summary, code, image, or other artifact generated by the AI System in response to Customer Data or prompts. 1.4 **"Training Data"** means data used to train, fine-tune, adapt, or otherwise improve an AI System's parameters, weights, or behavior. 1.5 **"High-Risk AI Use Case"** means use of the AI System in any application that materially affects individuals' access to credit, employment, housing, education, healthcare, essential services, or that constitutes a use case regulated as "high-risk" under applicable AI law, including without limitation the EU AI Act. 1.6 **"Applicable AI Law"** means any statute, regulation, executive order, or binding guidance specifically governing artificial intelligence, automated decision-making, or algorithmic systems, including the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), the Colorado AI Act (C.R.S. § 6-1-1701 et seq.), and any similar law in force in a jurisdiction where Customer operates. --- ## 2. Training Data Restrictions 2.1 **No Training on Customer Data.** Vendor shall not use Customer Data as Training Data to train, retrain, fine-tune, or otherwise update any AI System's parameters unless Customer provides prior written consent specifying: (a) the specific AI System to be trained; (b) the categories of Customer Data to be used; and (c) the purpose and scope of training. 2.2 **Isolation Obligation.** Where Customer Data is used for fine-tuning pursuant to Section 2.1, Vendor shall maintain logical and technical isolation such that fine-tuned weights derived primarily from Customer Data are not commingled with weights serving other customers. 2.3 **Third-Party Foundation Models.** If the AI System incorporates a third-party foundation model (e.g., a model licensed from a third-party AI developer), Vendor shall disclose: (a) the identity of the model provider; (b) the training-data practices of that provider as disclosed in the provider's published data card or acceptable-use policy; and (c) any known restrictions on commercial use of model Outputs. 2.4 **Data Retention for Training Artifacts.** Any intermediate Training Data copies or embeddings derived from Customer Data shall be deleted within ___________ days of the completion of the relevant training run, unless longer retention is required by law. ---