An AI apocalypse? Hong Kong’s real challenge is much less dramatic

Hong Kong's primary concern regarding artificial intelligence (AI) is not the dramatic notion of an AI apocalypse, but rather the more subtle risks that accompany the rapid advancement of the technology. As governments, companies and investors scramble to adapt to this new era, there is a growing gap between the excitement surrounding AI and the need to understand the potential consequences.

While investors may be willing to tolerate uncertainty for the sake of financial returns, regulators must take a more cautious approach. Companies rushing to adopt AI solutions without fully understanding their implications may inadvertently import the problems of other organisations. This highlights the importance of government guidance to ensure that businesses use AI safely and responsibly.

The risks are not limited to bad actors; even those without malicious intent can be fooled by AI tools, leading to errors that can erode trust in the system. A recent case in Hong Kong revealed how identity fraud was committed using AI-generated documents, and unrepresented litigants were making mistakes due to the assistance of generative AI.

Although the mistake was not intentional, the potential for misinformation is a real concern. The danger lies not only in the machine making a wrong decision, but in the misuse of the tool by users who lack the necessary expertise. Hong Kong's government is addressing these concerns by examining the legal implications of AI and reviewing relevant legislation.

The Department of Justice is spearheading an interdepartmental review to support the wider use of AI while also addressing issues such as AI-enabled crime, liability and the management of AI agents. This is a positive step, but it is equally important to equip people with the knowledge and skills needed to understand and use AI effectively.

Ultimately, the challenge is not to treat every application of AI as dangerous, but rather to ensure that the technology is used responsibly, particularly in high-stakes domains where the consequences of errors can be significant.