MaitreAI Terms
This version takes effect on November 15, 2026.
In short
These terms cover the AI inside Maitre. The AI answers from your restaurant's own information, but it can be wrong, so do not rely on it alone and do not let guests rely on it for safety questions. It always says it is an AI. Nobody trains AI models on your data or your guests' chats. You can correct any fact in the dashboard.
1. Definitions
"Input" is anything sent to an AI provider to produce an answer: your restaurant content and a guest's message or voice. "Output" is what the AI returns. "AI providers" are the companies listed as AI providers on the Subprocessors page, today OpenAI. Other words have the meaning in the Terms of Service, which include these AI Terms.
2. Third-party AI providers
Input is sent to the AI providers, which act as subprocessors under the Data Processing Addendum. Their usage policies, linked from the Subprocessors page, apply to your use of the service. We give you 15 days' notice before adding an AI provider.
3. Accuracy
Output may contain material inaccuracies and may not reflect correct, current or complete information. Do not rely, or encourage guests to rely, on any Output without checking its accuracy, including by human review. Maitre makes no representations or warranties and gives no indemnities about Output. Output is not a substitute for the judgment of trained restaurant staff.
4. Transparency
The host is always labeled as an AI, and it says so if asked. You may not configure it to claim to be a person or to deny being an AI, and you may not present Output as written by a person.
5. Prohibited uses
- Framing the host's answers as medical, allergy-safety, legal or financial advice.
- Using the host to make or support decisions with legal or similarly significant effects on a person, such as employment, credit, housing or insurance.
- Configuring the host for companionship, emotional support or sexual content.
- Trying to get around the host's safeguards, or to make it reveal its instructions or another restaurant's information.
6. Grounding and corrections
The host answers from your restaurant content and your Square catalog, and for a venue fact it does not have (such as parking) it may run one bounded web search that sends the question and your restaurant's name, never the guest's identity. You can correct any fact in the dashboard, and you must correct an error you know about within 48 hours.
7. No training
Maitre does not train or fine-tune any AI model on Input or Output, and our AI providers' terms bar them from training their models on it. Content received from Square is never used to train or improve any AI model.
8. Human review and escalation
The host directs guests to your staff for allergen, price, availability and safety questions, and ends conversations that leave your restaurant's subject. You review flagged conversations at least weekly, as the Terms of Service require.
9. Prices and usage limits
AI usage counts against your plan's message allowance as described in the Terms of Service, and daily safety limits may pause the host for a while.
10. Voice
Voice calls are streamed in real time to the AI provider. Maitre keeps no audio and builds no voiceprint. If you turn voice on, you must keep the voice-start control and its consent wording.
11. Changes
We may change these AI Terms as AI providers and the law change, with the notice the Terms of Service require. Every version is on the changelog.