AI Pitfalls: Accountability and Product Liability

Hello, this is Hirose.

In the modern business landscape, generative AI is becoming an indispensable tool. From gathering information and generating ideas to analyzing complex data, it dramatically streamlines our work, much like a highly capable assistant. As Professor Karim Lakhani of Harvard Business School stated, "Just as the internet dramatically lowered the cost of information transmission, AI will lower the cost of 'thinking'," AI is fundamentally changing the way we work.

However, hidden behind this convenience are significant risks that cannot be overlooked. If AI output leads to unexpected results or incorrect information, are we prepared to answer the questions, "Why did this happen?", and "Who bears the responsibility?"?

In this article, we will delve into two critical challenges faced by companies utilizing AI: "accountability" and "product liability".

1. 'Accountability' Hidden in Inexplicable Behavior

AI, especially generative AI, tends to be a "black box" where the output results are opaque. With traditional software, one could identify the cause of a defect by analyzing the code, but because generative AI involves complex algorithms and vast amounts of training data, it is extremely difficult to clearly explain why a specific output was reached.

The inability to answer this "why?" is what undermines trust in AI.

AI developers are required to have "accountability" to explain the basis and process of their output results. Fulfilling this responsibility is essential for enhancing the reliability of AI and ensuring that users can continue to use the service with peace of mind.

2. 'Product Liability' to Prepare for Unexpected Risks

The risk that AI might output incorrect information or discriminatory expressions, causing harm to users, cannot be ignored. If such a situation occurs, who should bear the responsibility?

This is where the "AI developer's "product liability" is called into question. AI is considered a "product" that can potentially cause disadvantage to users through its output. Therefore, developers have a responsibility to predict the potential risks posed by AI and take appropriate measures to minimize them.

In the unlikely event that a user suffers damage due to inappropriate AI behavior, the developer may be held liable.

3. A Sincere Request to AI System Development Companies

Recently, there has been an increase in young people calling themselves "AI engineers" simply because they can use AI-related APIs. While some might call them "fake AI engineers," can such companies explain the complex behavior of AI systems or take appropriate responsibility when defects occur?

Irresponsible responses like the following lose customer trust and ultimately lead to distrust in AI technology itself.

Even if they are using an external API, as long as they are releasing that AI system into the world, the development company must take responsibility for its output results. They must be prepared to respond sincerely to customer inquiries and fulfill their "accountability" and "product liability".

4. A Warning to Companies Introducing AI Systems

For companies introducing AI systems, this is not someone else's problem.

When introducing an industry-specific AI system developed by an SIer, it is essential to verify whether that development company possesses sufficient "accountability" and "product liability capability".

• Accountability
  Can the rationale and processes behind an AI system's output be clearly explained? Without this, adopting companies cannot understand AI behavior, making appropriate utilization and improvement difficult.

• Product Liability
  Can the developing company take responsibility for damages caused by defects or malfunctions? Without this, adopting companies face the risk of suffering significant losses in the event of an emergency.

While the introduction of AI systems contributes to operational efficiency and decision-making support, it also carries significant risks. Do not proceed with implementation lightly; instead, rigorously evaluate the capabilities of the developing company to ensure safe and effective AI utilization.

5. Conclusion

Generative AI has the potential to significantly transform our society, but to realize its true value, it is essential to always be conscious of the two critical aspects of "accountability" and "product liability" and to build an appropriate relationship with AI.

By having AI developers, AI users, and companies adopting AI systems all understand these responsibilities and engage with them sincerely, we will be able to build a society that maximizes the benefits of AI.

The content discussed here is based on insights I previously learned in the "Artificial Intelligence Implications for Business Strategy" course at the MIT Sloan School of Management.

Thank you for reading until the end.