AI Usage Policy

Version: 1.0  ·  Effective Date: March 23, 2026

This Policy is part of the Terms of Service and regulates the use of AI systems on the Tyfoon platform.

1.1. The Tyfoon platform uses AI systems for:

• automated code review;
• validation of answers to assessment tasks;
• generation of summaries of mentorship consultations.

1.2. In accordance with the Law of the Republic of Kazakhstan "On Artificial Intelligence," the AI systems used are classified as medium-risk systems.

1.3. The Platform uses language models provided by third parties (OpenAI, Google) and speech analysis services (Fireflies.ai).

2.1. Each result generated or modified by AI algorithms is accompanied by a visual label "Verified by AI" or a similar notification.

2.2. The User acknowledges that AI responses are based on statistical patterns and are probabilistic in nature. The Provider does not guarantee the absolute accuracy of each AI comment.

3.1. The Provider makes reasonable efforts to ensure that the results of the AI systems comply with applicable law and conducts regular assessments of AI security and reliability risks.

3.2. The Provider is not responsible for AI inaccuracies arising from the probabilistic nature of the algorithms, changes in the models of third-party providers, or misuse of AI by the User.

3.3. Limitation of the Provider's liability is governed by Section 7 of the Terms of Service.

4.1. The User has the right to request an explanation of the logic and principles of operation of AI algorithms if the automated decision significantly affects their assessment or access to further course materials. Requests are sent to support@tyfoon.kz.

4.2. AI decisions that affect the User's final assessment or the issuance of a certificate are subject to review by a human mentor upon the User's written request. The review is carried out within 5 working days. This right does not apply to intermediate assessments, homework assignments, and practice test results.

5.1. The use of the User's program code for further training of the Platform's AI models is permitted only in anonymized form and in the absence of a direct prohibition by the User in the account settings.

5.2. Withdrawal of consent to the use of data for training does not entail the deletion of information from already trained models, since anonymized data cannot be identified and extracted from the trained model.

The User is prohibited from:

• applying "prompt injection" methods to bypass security restrictions;
• attempting to extract the source code, weights, or architecture of the Platform's AI models;
• intentionally manipulating the outputs of AI systems in order to obtain unfair assessments or privileges;
• using the Platform's AI systems for adversarial purposes, including the creation of adversarial examples;
• using the data obtained from the Platform's AI systems to create competing datasets or train third-party models.

For questions regarding AI functionality and to challenge decisions: support@tyfoon.kz