With AI, It’s All About the Data

Going beyond the basics of artificial intelligence (AI), it is important to focus on the data. How data is used, managed, and understood in the context of AI has become the center of many legal and business conversations, particularly as AI technologies are increasingly integrated into products and services. In this environment, data is not just a resource, it is the lifeblood of AI, influencing everything from product development to risk mitigation.

EARLY DAYS OF AI INTEGRATION

Initial Concerns

The focus originally was education—how AI works, its potential risks, and its impact on business processes. Legal teams and IT departments were especially concerned with understanding the technology's limitations and risks, such as biases in data processing and potential compliance issues.

Development of AI Usage Policies

As businesses began to adopt AI, the need for formal AI usage policies emerged. These policies helped organizations regulate how AI could be used, setting boundaries, especially in customer data processing.

Shifting Focus to Use Cases

Over time, AI projects began to transition from high-level risk assessments to more practical applications, such as framing specific use cases for AI and negotiating the terms for integrating AI into business processes. Companies were focusing on how to use AI effectively and responsibly, including understanding what AI models (like large language models, or LLMs) can and cannot do.

From Risk to Negotiation

The focus shifted from simply advising businesses about risks to negotiating deals that define data usage rights and responsibilities. Negotiations around AI contracts began to revolve around how data would be used, how AI models would process it, and who would be accountable if something went wrong.

KEY CONTRACT CONSIDERATIONS

With the growing reliance on AI, it is important that businesses address contractual issues to protect their interests and manage risk.

Responsibility to Test AI

Of the most pressing issues companies face is ensuring that AI models are tested for accuracy and reliability. Contracts should include provisions that hold the vendor accountable for testing AI to ensure it meets the following criteria:

• Generates accurate and reliable data
• Avoids biases in how data is used or the output it creates
• Identifies and addresses defects in AI models
• Complies with legal and regulatory standards

Customer Concerns

Customers are focused on understanding what data is used in AI models, how it is processed, and whether the output meets specific standards. If the AI system uses biased or flawed data, the customer could face legal and reputational risks. Therefore, contracts for products or services that incorporate AI should be explicit about data rights, transparency, and accuracy.

Vendor Considerations

Vendors may already have their own policies or principles, whether published or provided during negotiations. Customers should ensure that these policies align with their standards and expectations.

THE ROAD AHEAD: REGULATORY AND ETHICAL CONSIDERATIONS

The landscape of AI is rapidly evolving, and regulatory frameworks are starting to catch up. Jurisdictions worldwide, including those in both the United States and European Union, are introducing laws and frameworks.

Regulatory Frameworks

In the US, federal and state regulations are increasingly focused on the role of AI in business decisions and operations. Similarly, the EU AI Act and regulations in the United Kingdom emphasize the need for businesses to understand the implications of AI in decision-making processes.

Key Principles of AI

The key principles of AI include the following:

• Safety, security, and robustness: AI systems must be built to withstand threats and recover from incidents
• Transparency and explainability: It should be clear how AI systems make decisions, especially in high-stakes environments
• Fairness: AI should avoid discriminating against individuals or groups
• Cost-benefit considerations: Organizations must weigh the costs of implementing AI against its potential benefits, ensuring that they are not just adopting AI for the sake of innovation, but for tangible improvements

THE AI PROCESS: KEY STEPS AND RISK MANAGEMENT

Implementing AI successfully requires careful planning, risk assessments, and continuous monitoring, all tailored to the specific needs of the business.

Some key steps to manage risks include the following:

• Identifying AI applications: Begin by understanding the different areas within the business where AI can be applied, from process automation to decision-making
• Conducting AI risk assessments: Businesses should assess potential risks, including:
  • Use cases: Define the specific tasks or problems that AI will address
  • Data accuracy and verifiability: Ensure that the data used is of high quality and accurate
  • Vendor lock-in: Consider whether the company could become overly reliant on a single AI vendor, potentially limiting flexibility
• Integrating AI policies: AI policies should be integrated into business operations; a risk-based approach should guide decision-making, with categories for unacceptable, high, limited, or minimal risks
• Assurance techniques: Regular assessments, audits, and certifications help verify the integrity of AI systems and processes
• Shift in focus: Businesses should also shift their focus from just safeguarding against security threats to ensuring resilience—that their AI systems can recover from incidents
• Third-party providers: As AI increasingly relies on third-party providers, companies must ensure that these providers comply with the same resiliency policies and provisions; the ability to monitor and verify third-party compliance is critical

KEY TAKEAWAYS

Businesses need to understand that AI is no longer just a concept, but rather a central component of modern operations. The way data is used and managed in AI systems will play a critical role in ensuring success. Businesses should consider the following takeaways:

• New era: We are moving beyond just risk assessments to active deal-making around AI integrations
• Data reigns supreme: Data processing and consumption are directly tied to pricing structures, and data usage rights are more important than ever
• Contract negotiations are taking shape: The landscape of AI deal negotiations is evolving, with a focus on pricing, testing responsibilities, data usage rights, and risk allocation
• Moving forward: A principles-based approach that is contextual, cross-functional, and resilient will be essential for businesses as they navigate AI's role in their operations. And at the heart of all these changes, data remains king