The burgeoning field of artificial intelligence (AI) presents a plethora of opportunities for businesses. Yet, as organisations increasingly turn to AI systems to streamline operations, enhance decision-making, and improve customer experiences, they must also grapple with the unique challenges of contracting for AI projects. This blog post aims to shed light on some of the key considerations when engaging with AI providers, ensuring that businesses can leverage this powerful technology while safeguarding their interests and minimising potential risks.

Data serves as the lifeblood of AI systems, making it a central point of consideration in any AI project contract. Organisations must carefully define the following aspects of data usage:

• Data Access: clearly delineate which documents and data the AI system will be granted access to. Define clear boundaries to protect sensitive information and ensure compliance with data protection regulations.

• Data Provenance: establish the origin of the data and ensure the relevant parties possess the necessary rights for its use. This step helps prevent legal disputes and copyright infringement issues.

• Data Quality: determine if the data requires cleaning or preprocessing and who will be responsible for this task. Clean, high-quality data is critical for training accurate and effective AI models.

• Data Conversion: address the need for data conversion into the supplier’s proprietary formats, if applicable. Establish clear responsibilities and costs associated with data conversion.

• Personal Data: if the AI system will process personal data, ensure compliance with relevant data protection laws, such as the GDPR. Obtain necessary consents and implement appropriate safeguards.

• Confidentiality: establish robust confidentiality clauses to protect sensitive information and intellectual property, especially in the context of know-how and AI model training.

Another critical aspect of AI project contracting involves managing and utilising the results generated by the AI system. Considerations include:

• Presentation of Results: define how the output of the AI system will be presented. Specify the format (e.g., reports, visualisations) and level of detail required for meaningful insights.

• New Know-How: address ownership and usage rights of any new knowledge or insights generated by the AI system. Clarify whether these belong to the customer or the AI provider.

• Customer-Specific Customisations: determine ownership and licensing terms for any customisations or enhancements made to the AI system to meet the customer’s unique requirements.

• Usage Data: specify ownership and permissible use of data generated from the customer’s interactions with the AI system. This data may have significant value for further AI development and improvement.

While the standard obligation for AI providers to perform services with “reasonable skill and care” is important, AI project contracts often require additional details regarding service quality and standards:

• Governance and Verification: define responsibilities for governing and verifying the outputs of the AI system. Establish an objective and measurable framework for identifying, rectifying, and mitigating any errors or biases.

• System Interoperability: if the AI service interacts with other parts of the customer’s system, clearly specify those interactions and define the customer’s expectations regarding compatibility and data exchange.

• AI Service Selection: if the supplier is selecting an AI service for the customer, ensure clarity about the model’s provenance, including details about its developers, training data, intended and prohibited uses, and licensing terms.

• Model Training: if the supplier is responsible for training the AI model, define clear criteria for evaluating the training process and outcomes. Specify metrics for assessing model accuracy, fairness, and robustness.

• Regulatory Compliance: if the customer operates in a regulated environment, ensure that the AI services enable adherence to all relevant regulatory requirements.

AI project contracts should also address governance issues to ensure proper control and oversight of the AI system:

• Data Retention: define policies for document and data retention, including logging practices for recording the AI system’s operations and decision-making processes.

• Audit Rights: establish audit rights for the customer, allowing for first-party or third-party audits to verify compliance, data usage, and model performance.

As AI-specific regulations are enacted and come into force, AI project contracts will need to account for these evolving requirements:

• Horizon-Scanning: assess the likelihood of the proposed AI use case falling within the scope of any upcoming AI regulations.

• Material Impact: if regulatory compliance is likely, estimate the potential impact on the project, including costs and operational changes.

• Contractual Flexibility: if regulatory changes are anticipated, consider including provisions for renegotiating terms, termination rights, or mechanisms to address potential commercial impacts.

• Compliance Lead Time and Costs: factor in the likely lead time, processes, and costs involved in bringing the AI system into compliance with new regulations.

Contracting for an AI project demands careful attention to a range of issues, from data management and output utilisation to service quality, governance, and regulatory compliance. By proactively addressing these considerations in contracts, businesses can harness the transformative power of AI while protecting their interests and ensuring long-term success in this dynamic technological landscape.