The legal advice that AI and fintech startups need

Photo credit: Sergey Soldatov / 123RF

This article is part of Tech in Asia’s partnership with Asia Law Network where we publish articles written by lawyers on their advice for startup founders. For more articles in this series, go here.

Previously, we shared with you some legal advice that startups in ecommerce, sharing economy, and VR need to know.

In this article, a couple of professionals share some legal insights (based on Singapore’s legal system) that detail the importance of regulations and compliance in the fields of AI and fintech.

Artificial intelligence

By Samuel Ng, counsel at BlackOak LLC

AI is not yet fully regulated, but existing legal concepts can still apply to its problems—at least for the most obvious areas of concern.

One such area is that of data collection and usage. AI relies on big data to make many types of predictions. But the truth is that once a machine is taught to learn, it can be difficult to comprehend why it reacts a certain way to a particular set of data inputs. These mystifying inner workings of the AI are the reason machine learning models are referred to as “black boxes.”

In the face of such opaqueness and the trial-and-error nature of big data analytics, can any binary yes-or-no consent given by consumers still truly be meaningful and compliant with existing personal data regulations?

A possible solution, as sometimes raised in other circles, may lie in graduated consent. Instead of a binary option at the outset, consumers are given the option to agree or disagree to various uses of their personal data. But for the most part, it may just be prudent to have a comprehensive, well-drafted terms of use, and be upfront with the exact known purposes for which data is being collected, stored, and used.

Who should bear the blame? Who should be liable for decisions that AI makes without human intervention?

Another issue is that of liability. For example, in October 2016, while out on a test drive, a self-driving car got into a collision with a lorry. Although no one was hurt in this experiment, the risk of autonomous ambiguity resulting in human injury (or even death) remains very real. Who should bear the blame? Who should be liable for decisions that AI makes without human intervention?

Perhaps, the situation can be more complex as the insurance industry is just about to play catch up. But between a business peddling AI and its customers, liability may in the meantime be managed through a good understanding of contract law and the scope of exclusion clauses. After all, a significant chunk of contract law is about the allocation of risk in business relationships, and many legal pitfalls occur because of a lack of foresight and planning.

Closely related to this is the ethics of machine learning. Machine learning can produce decisions that may be strategically correct but ethically wrong. In 2017, we learned with horror through Microsoft’s Twitter chatbot, Tay, that machines can turn racist, sexist, and otherwise prejudiced in less than 24 hours by picking up the same tendencies in human data inputs.

Fintech