Hong Kong: Responsible and Sustainable AI use in Employment

In the first of our series on ‘responsible AI use’ we will discuss two common uses for AI in HR management: resume screening and learning and development. Whilst using AI tools in resume screening can save time there is a risk that they can contain and magnify biases, including those based on protected characteristics. Also, using AI to personalize learning and development plans helps to support talent acquisition and retention, but requires gathering and storing employee data, so is subject to data protection laws. 

Whilst we see the rules around AI evolve, companies are advised to align their AI usage with the common principles of fairness, transparency and privacy, which are prevalent in current legislation and guidance. This alignment not only ensures ethical AI use but also prepares companies for the potential shift from guidance to mandatory compliance.

Potential bias in resume screening

The use of AI in recruitment processes offers the potential for automation and simplification. In the context of resume screening, AI has several uses:

1. Screening tools: the use of natural language processing to evaluate CVs and personal statements and assign a score to applicants based on keywords
2. CV matching: tools that match applicants with ideal CVs based on semantic similarities
3. AI-powered psychometric tests and games: used to measure cognitive skills and job-relevant personality traits

These tools can process many resumes within minutes but they can also contain and magnify biases embedded in the data they are trained on, including those based on protected characteristics. Companies should be aware that, under Hong Kong’s’ four anti-discrimination ordinances, a total of seven protected attributes are covered (sex, pregnancy, marital status, disability, family status, race and breastfeeding). If AI-powered tools screen candidates based on the above criteria, the employer may be subject to a claim of unlawful discrimination.

For screening tools trained based on historic data, some of the data used may be drawn from prior successful candidates, with a view to selecting similar candidates in future recruitment exercises. Insofar as the selection criteria relates to the seven specific protected attributes, this could potentially attract ‘algorithmic discrimination’ resulting in the underrepresented groups continuing to be underrepresented in the workplace. If the screening tools specifically take into account any particular protected attributes then this could potentially give rise to a claim for direct discrimination.

Further, where the employer applies a criterion in the screening process which has the indirect effect of placing individuals with a certain protected attribute at disadvantage and that criterion is not justifiable, this may amount to unlawful indirect discrimination.   

To mitigate these risks, companies should train AI on diverse data, regularly monitoring and auditing systems for biases, and stress test the software before full implementation.

Data protection obligations with personalized training

AI in learning and development can provide personalized training based on individual needs. The expectation is that this will enhance the work environment and in doing so help support talent acquisition and retention. AI can also be used to predict future training needs through identifying skill gaps.

The gathering and storing of employee data to make such learning and development decisions about employees, which is used to train and fine-tune the AI system, is inevitably subject to data protection laws. In Hong Kong, the Personal Data (Privacy) Ordinance (Chapter 486 of the Laws of Hong Kong) mandates notification or consent obligations when data is collected, detailing how it will be used and to whom it will be transferred. In addition, on June 11, 2024, Hong Kong’s Office of the Privacy Commissioner for Personal Data published the Artificial Intelligence: Model Personal Data Protection Framework as a best practice guidance for organizations procuring and using AI solutions.

How are other jurisdictions managing responsible AI use in HR?

Hong Kongs mixture of guidance alongside existing data protection laws is similar to what we are seeing across the globe. We have seen that both the UK and US have published guidance on AI’s use in hiring and employment practices. Going further, four states in the US have enacted their own AI specific legislation in relation to HR practices. There is also the EU AI Act which came into force on 1 August 2024 and is the world’s first comprehensive framework on AI. In addition, the Global AI Treaty, aims to act as a bridge between jurisdictions on globally agreed AI governance standards.

As regulator guidance and laws evolve across the globe, companies are navigating a transition from non-binding guidance to mandatory compliance. As the adoption of AI models increase and the number of AI use cases expands, we anticipate a shift towards mandatory compliance in an increasing number of jurisdictions, potentially drawing inspiration from the EU model.

Co-authored by Nathan Handoll and Clare Johnston (Knowledge)

Further reading on the role of AI in HR processes

How AI Treaty Will Further Global Governance Cooperation | Artificial Intelligence

Singapore – New AI guide on synthetic data generation proposed

New Generative AI Governance Framework in Singapore

Global AI Regulatory Update - June 2024

EU AI Act: Parliament adopts groundbreaking regulatory framework