Artificial Intelligence Addendum (AI Addendum)

This Addendum applies to all AI-driven capabilities provided across Satia, including:

• Satia Docs & RAG: Semantic vectorization, contextual retrieval of Markdown files, PDFs, and computer vision captioning powered by Gemini 2.5 Flash.
• Autonomous Conversational Agents: Customer engagement and consultative agents deployed on WhatsApp, Instagram, WebChat, and email.
• Telephony Voice Agents: Model inference integrated with human voice synthesis (TTS/STT) and telephony carriers executing inbound and outbound calls.
• Dynamic ERP V3: Automated entity actions guided by semantic agent instructions within dynamic table schemas.

Our architecture operates under enterprise API agreements with Google Cloud (Gemini API) and OpenAI, ensuring strict zero data retention policies for training purposes (*Zero Data Retention for Training*). Customer data is processed ephemerally during inference and remains strictly isolated within your tenant.

Input Ownership: Customer retains exclusive title and all intellectual property rights in and to all data, documents, table parameters, transcripts, and prompts provided to the Service ('Inputs').

Assignment of Output Rights: As between Satia and Customer, to the maximum extent permitted by applicable intellectual property law, Satia hereby assigns and transfers to Customer all right, title, and interest in and to the responses, generated texts, and automations produced for Customer by the AI ('Outputs').

Non-Exclusive Outputs: Customer acknowledges that, due to the probabilistic mathematical nature of language models, Outputs generated for Customer may be identical or substantially similar to Outputs generated for other users submitting similar prompts. Such similarity shall not constitute an intellectual property infringement or breach of this agreement.

Customer expressly acknowledges and agrees that:

• Probabilistic Output: Generative AI operates via token prediction algorithms rather than deterministic calculation. Outputs may occasionally contain factual errors, logical inconsistencies, or unsupported information (industry-recognized 'hallucinations').
• No Guarantee of Accuracy: Satia does not warrant that AI Outputs will be error-free, complete, or fit for any critical decision-making purpose.
• Mandatory Human Oversight (HITL): For any consequential or regulated use cases (such as medical assessment, legal advice, credit underwriting, hiring decisions, or emergency dispatch), Customer is strictly obligated to implement human oversight before any decision produces binding effects.

Deploying automated telephony voice agents entails the following mandatory obligations for Customer:

• Mandatory AI Identification: Every call must include an immediate, audible disclosure informing the participant that they are speaking with an artificial intelligence assistant (e.g., 'Hello, I am an AI assistant calling on behalf of [Company Name]...').
• Call Recording Consent: Customer bears sole statutory liability for complying with wiretapping and call recording statutes across federal, international, and all-party consent states (e.g., California, Delaware, Florida, Maryland, Pennsylvania), securing prior express consent before recording or transcribing any call.
• TCPA & Telemarketing Compliance: Customer is prohibited from using Satia voice agents for automated telemarketing without required prior written consent, calling numbers on Do-Not-Call registries, or transmitting altered caller-ID data (*spoofing*).

Customer agrees not to use Satia's AI systems for:

• Generating child sexual abuse material (CSAM).
• Synthesizing non-consensual deepfakes to defame individuals or execute financial fraud.
• Generating malicious exploits, cyberattack scripts, or automated vulnerability probes.
• Providing unlicensed medical, legal, or financial professional advice.
• Attempting to bypass safety filters through prompt injection attacks or jailbreaks.

Satia integrates advanced foundational models provided by partners including Google LLC (Gemini 2.5 Flash / Pro) and OpenAI LLC.

Customer acknowledges that model providers may deprecate, upgrade, or tune underlying model checkpoints periodically. Satia commits to maintaining operational compatibility and routing requests to equal or superior models without material capability regressions.

SATIA LLC SHALL NOT BE LIABLE TO CUSTOMER OR ANY THIRD PARTY FOR ANY DIRECT, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING FROM:

• Commercial transactions or contractual commitments entered into by customer-configured autonomous AI agents.
• External LLM provider API outages or latency beyond Satia's direct engineering control.
• Statutory telecommunications fines resulting from Customer's failure to satisfy AI voice disclosure or call recording consent mandates.