Annex to the general terms and conditions: specific conditions for AI models

1. Data processing and "zero data retention"

1.1. Training prohibition: Swiftask and its AI model providers (OpenAI, Anthropic, Mistral, Google, etc.) are strictly prohibited from using the Client's Data (Inputs/Outputs) to train or improve their foundation models.

1.2. "zero data retention" (ZDR) policy: Swiftask uses the commercial ("Enterprise") APIs of its Providers. The Client's prompts and outputs are not retained by these third parties, are not used for learning, and are not subject to any human review by them.

2. Prohibited activities and compliance

The Client is expressly prohibited from using Swiftask to:

• Develop competing AI models (use of outputs for model training).

• Generate content intended to deceive, manipulate, or produce unsolicited mail (Spam).

• Attempt to bypass security filters (Prompt Injection). The client undertakes to comply with the applicable usage policies of each AI provider they choose to activate via the Platform.
  

3. Disclaimer of warranty

AI systems are probabilistic and may generate incorrect information ("hallucinations"). Use of these services is at the Client's risk, who has the obligation to verify independently and humanly the accuracy of Outputs before using them.