FALA AI Services Terms

Last updated: October 16, 2026

Additional terms for FALA’s AI features: how they work, their limits, disclosure duties, and how AI data is handled. These terms are part of the Terms of Service.

1. Scope

These AI Services Terms apply to every feature of FALA that uses artificial intelligence, including the visitor-facing chat and voice agent, knowledge base building and search, page guidance and form filling, the Architect, configuration suggestions, insights, summaries, image generation, and features available to AI assistants connected through our MCP server. Capitalized terms have the meanings given in the Terms of Service.

2. How FALA’s AI Works

FALA builds a knowledge base from the website pages, files, and answers you provide, and retrieves relevant passages to ground each response. Responses, speech, and transcripts are generated by AI models from third-party providers (currently Google, xAI, and OpenAI). We may change models, providers, voices, and engines at any time, and different requests may be served by different models.

When a Visitor asks for help on a page, the widget reads the content and structure of that page so the agent can guide the Visitor, highlight controls, and fill form fields the Visitor dictates.

3. AI Disclosure

FALA agents are designed to identify themselves as AI assistants and to disclose that they are AI when asked. You must not configure, instruct, or prompt an agent to claim to be a human, to deny being an AI, or to conceal that it is an AI system, and you must not remove or obscure the AI disclosure that FALA displays.

You are responsible for any additional disclosure that applies to your industry, location, or use case, including disclosures required for AI-generated or artificial voices on calls, for recording and transcription, and under laws such as Article 50 of the EU Artificial Intelligence Act and state AI and chatbot disclosure laws.

Voice and chat interactions may be sent to AI model providers for real-time processing, grounding, transcription, and response generation.

4. Limitations of AI

AI systems are probabilistic. Output can be wrong, incomplete, outdated, inconsistent, biased, or inappropriate, even when your knowledge base is correct, and the same question can receive different answers. Speech recognition can mishear names, numbers, email addresses, and other details. The agent may misunderstand a Visitor, fail to follow your instructions, or take an action you did not intend.

You are responsible for testing your agent before and after it goes live, monitoring its conversations, correcting your knowledge base and briefing, and using FALA’s controls (such as claim corrections, staged knowledge, and publishing) to keep it accurate.

5. Human Review and Prohibited High-Risk Uses

Do not rely on AI output for legal, medical, financial, emergency, or other high-risk decisions without qualified human review.

You must not use FALA’s AI to make or materially support decisions that produce legal or similarly significant effects on individuals, to provide professional advice, to handle emergencies, to interact with children, or for any use that would be classified as high-risk or prohibited under the EU Artificial Intelligence Act or similar laws, unless we have expressly agreed in writing.

6. Page Guidance and Form Filling

The agent fills fields only with information the Visitor provides in the conversation, never reads or fills password or payment fields, never ticks consent or terms checkboxes, and never submits a form; the Visitor reviews and submits it. You are responsible for your forms, their labels, the consents they collect, and what happens after a Visitor submits them.

7. Knowledge and Website Reading

When you give FALA a website address, FALA reads that site’s pages, honoring its robots directives, to build your knowledge base. You must own or be authorized to use that content. Automatic re-reads may bring back content you removed from FALA but not from your website, so keep your website and your corrections in sync.

8. AI Data Handling

Elav8 does not use your Customer Content or Visitor conversations to train or fine-tune AI models, and uses its providers’ business API services, under which they do not use that content to train their models. Providers may keep prompts and responses for a limited period, typically up to 30 days, solely to detect abuse and comply with law. Our Privacy Policy and Data Processing Addendum describe our handling of personal data in more detail.

9. Voices, Names, and Likeness

FALA’s voices are synthetic. You must not use FALA to imitate the voice, name, or likeness of a real person, or to create content that a reasonable person would mistake for a real person’s statement, without that person’s documented consent and every disclosure the law requires.

10. Connected AI Assistants

If you connect an AI assistant or agent to your workspace through our MCP server or an API key, it acts as one of your Users. Every action it takes is your action under the Terms of Service, and you are responsible for the permissions you grant it and for disconnecting it when no longer needed. Content it reads from websites, knowledge bases, or Visitor messages is untrusted input and may contain instructions it should not follow.

11. Safety Measures

We and our providers may filter, block, or modify inputs or Output, refuse requests, limit features, and suspend agents to prevent harm, abuse, or legal violations. These measures are imperfect, and their presence does not shift responsibility for your agent’s conduct to us.

12. Output and Warranty

Ownership of Output, and the disclaimers, limitations of liability, and indemnities that apply to AI features, are set out in the Terms of Service. AI features are provided as-is, without any warranty of accuracy, completeness, or fitness for a particular purpose.