Artificial Intelligence 101: I Use AI, What Should I Consider in Contracts?

Artificial intelligence (AI) has become one of the cornerstones of digital transformation through big data analysis, automation, and decision support systems. However, the inherent nature of AI, its continuous learning and dynamic evolution, raises critical legal questions, particularly regarding regulations and contract provisions. To address these questions, it is first essential to understand how AI operates.

AI can generally be defined as systems that can perform functions typically requiring human intelligence, such as perception, learning, communication, creativity, decision-making, reasoning, and development. Most AI models consist of four fundamental stages:

Each of these stages introduces considerations that must be addressed in contracts. Key issues include data usage, reliability of outputs, and model updates, all of which necessitate clear definitions of the parties’ responsibilities.

Furthermore, the role of AI in a contract is crucial since it defines liability boundaries. AI can be used in various capacities, such as:

• - As an independent software tool (e.g., a Q&A system),
• - Integrated into a physical product (e.g., autonomous vehicle software),
• - Enhancing a physical product’s functionality without integration (e.g., AI-based medical imaging analysis systems),
• - As a component of a larger system (e.g., decision-support systems in smart manufacturing).

These variations significantly impact the structure, rights, and obligations in contracts between AI developers and service recipients. Consequently, different contract types, such as software development agreements, R&D agreements, licensing agreements, or service agreements, may be necessary.

Regardless of the contract type, when AI is involved, the following key points must be considered:

• - AI systems can make errors. They may produce “hallucinations-outputs” that appear correct but are actually incorrect. Therefore, contracts should clearly state that no guarantees are provided regarding the accuracy of AI decisions and explicitly define liabilities for incorrect results.
• - AI-generated content may infringe intellectual property rights. AI systems process uploaded data and generate outputs, potentially creating IP risks. Contracts should clarify which data can be uploaded, who owns the AI-generated outputs, and how IP-related risks are allocated. Additionally, detailed provisions on handling input and output-related IP infringement should be included.
• - AI often processes customer data for training purposes. Unauthorized use of such data could violate data protection laws like KVKK (Turkish Data Protection Law) and GDPR. Therefore, contracts must specify whether customer data can be used for training, whether customer consent is required, and under what conditions data usage is permitted. These issues should also be addressed through data processing agreements.
• - Confidentiality and data protection are critical. The scope of confidential information, especially regarding inputs and outputs, must be clearly defined. Additionally, contracts should explicitly state that AI cannot use confidential information beyond the scope of the provided service.
• - AI’s dynamic and evolving nature may create risks or advantages. As AI systems are updated, contract risks may change. If certain conditions are met, contracts should allow for adaptation to changing circumstances. For instance, an unforeseen technological breakthrough could necessitate contract modifications.
• - AI can be misused for unethical or illegal purposes. AI algorithms, such as those used in recruitment, may unintentionally introduce bias and discrimination, leading to legal challenges. Contracts should clearly define each party’s responsibilities regarding AI’s compliance with ethical and legal standards.

These are fundamental considerations for AI-related contracts, but additional critical points may arise depending on the business model, AI usage method, and objectives. Ultimately, in the fast-paced world of technology, it is important to recognize that using AI inherently involves a chain of contractual relationships.