No model training

AI Usage & Responsible Use Policy

Ethical standards, customer data isolation, and responsible AI guardrails governing our restaurant intelligence platform.

Effective
September 4, 2026
Entity
Antigravity
Jurisdiction
India
Applies to
kafei.in
Strict Zero-Training Guarantee on Customer Proprietary Data

Kafei NEVER uses your confidential recipes, proprietary dish ingredients, business financials, customer records, or Google OAuth account data to train, retrain, fine-tune, or improve publicly accessible or foundational AI models.

1. Purpose and Scope

Kafei integrates cutting-edge Artificial Intelligence (AI) and Machine Learning (ML) technologies—including integrations with Google Gemini models, time-series forecasting algorithms, and natural language copilot assistants—to optimize restaurant kitchen workflows, predict inventory exhaustion, streamline menu engineering, and enhance operational decision-making.

2. Core Ethical Principles

2.1 Transparency & Explainability

We believe hospitality operators must understand how AI recommendations are formed. Kafei provides contextual rationale for inventory restock alerts, demand surge forecasts, and menu modifier recommendations.

2.2 Human-in-the-Loop Oversight

AI in Kafei is strictly designed as an intelligence amplifier, not an autonomous replacement for human judgment. Critical operational decisions—including automated purchasing orders, price overrides, staff shifts, and refund authorizations—require explicit human confirmation.

2.3 Fairness & Non-Discrimination

Our algorithmic models are engineered to prevent discriminatory outputs based on protected attributes, ensuring equitable service recommendations and fair pricing models.

3. Enterprise API Privacy & Data Segregation

  • Enterprise Zero-Data-Retention: When Kafei connects with foundation model providers (such as Google Gemini APIs), all API interactions occur over secure, enterprise-grade endpoints subject to strict zero-data-retention terms where prompts and completions are not logged or used for model training by the provider.
  • Tenant Isolation in AI Memory: Contextual memory and embeddings used by the Kafei AI Copilot are logically segregated per tenant ID. No restaurant’s data is ever exposed to or accessible by another subscriber.

4. Specific AI Capabilities & Guidelines

4.1 Demand & Sales Forecasting

Evaluates historical POS velocity, seasonal trends, day-of-week patterns, and table turn rates to project daily ingredient demand. Chefs and kitchen managers should adjust predictions based on local weather, private events, or sudden market shifts.

4.2 Recipe & Menu Engineering Copilot

Analyzes margin percentages, ingredient wastage, and dish popularity (Stars, Plowhorses, Puzzles, Dogs) to recommend menu optimizations. Recommendations are advisory; subscribers retain complete discretion over dish formulations and retail pricing.

4.3 Kitchen Load Balancer & Prep Time Predictions

Computes dynamic prep times based on active KDS ticket volume and station workload. Dynamic cooking countdowns visible to diners on table QR screens reflect real-time kitchen state to set accurate guest expectations.

5. Prohibited AI Uses

Subscribers, operators, and staff are strictly prohibited from using Kafei AI tools to:

  • Generate misleading, fraudulent, or deceptive food descriptions or allergen disclosures.
  • Implement predatory surge pricing models that violate local consumer protection statutes.
  • Automate adverse employment termination decisions without human review.
  • Attempt prompt injection attacks, jailbreaks, or extraction of internal system prompts or other tenant data.

Questions about this policy?

Write to us for data subject requests, deletion requests, security reports, or compliance reviews. We respond within 5 business days.

Antigravity, Kolkata, West Bengal, India