Prof Dr Mohammad Ershadul Karim, a senior lecturer with University Malaya’s Law Faculty, points out that the same thinking applies to artificial intelligence (AI), which requires massive amounts of data to function.
He says laws relating to the collection and usage of personal data like Malaysia’s Personal Data Protection Act 2010 (PDPA) act as the first layer of defence in protecting public privacy in the face of widespread adoption of AI systems.
However, he adds that there are other aspects of AI not covered by such laws which have to be addressed, including accountability and transparency, to prevent bias and protect human rights.
He says one of the main concerns is the use of it for surveillance and crime prevention – for example, using a facial recognition AI to monitor CCTV (close-circuit TV) footage for suspicious individuals.
“In our common law system there’s the principle of natural justice, where a person can only be viewed as a criminal once found guilty by a court.
“But to use AI to decide if someone is a criminal based on body language and social information... who are they to judge?” he says, warning that misuse of such tech could endanger personal liberty.
In neighbouring Singapore, which installed over 90,000 CCTVs in public locations, a recent test using a robot to monitor “undesirable social behaviour” has drawn more privacy concerns.
Xavier – a wheeled robot that has seven onboard cameras – patrols residential and shopping areas looking out for people smoking in prohibited areas, improperly parked bicycles and those breaching coronavirus social-distancing rules.
Digital rights activists in the country remain unsatisfied even though the government has assured that Xavier is not being used to identify or punish offenders during the trial period.
In online reports, some pointed out that Singapore has made a U-turn on data privacy before.
In the case of its coronavirus contact-tracing app TraceTogether, the government initially assured that the data would only be used to curb the spread of Covid-19.
However, Singapore Minister of State in the Home Affairs Ministry, Desmond Tan, was reported saying that the police are empowered to obtain any data, including from TraceTogether for criminal investigations.
Lau Pin Lean, a lecturer in Bio-Law at the Law School of Brunel University London in the United Kingdom says it’s nearly impossible to opt-out of being watched by AI in a modern society.
For instance, e-commerce sites track users’ shopping habits in order to suggest more things for them to buy.
“It’s hard to escape. I use Amazon Prime because I can get what I want the next day.
“So that means surrendering data to the company or just not using its services,” she says, adding that in most cases people don’t really have a choice unless they are willing to forgo valuable services.
Laying down the law
The European Union (EU) in April proposed an Artificial Intelligence Act that will set a common regulatory and legal framework for the use of such technologies in its 27 member countries.
Lau applauded the move as the EU didn’t go for a simple one-size-fits-all solution, instead opting for a comprehensive approach that categorises AI based on risk.
The proposed law sorts AI into three broad classifications: prohibited or an unacceptable risk; high risk; and low or minimal risk.
AI systems in the top tier are unacceptable and must be banned as they pose a danger to people’s freedom and lives, says Lau, citing social scoring and subliminal systems as examples.
A social scoring system works like a credit scoring system but instead of measuring a person or company’s financial trustworthiness, it is used to monitor people’s behaviour.
The Social Credit System used in China has been criticised for infringing citizen’s rights to free speech as people have been penalised for airing dissenting views on the government.
Meanwhile, the EU defines subliminal systems as having significant potential to manipulate people through subliminal techniques to alter behaviour.
Lau found it interesting that during the dialogue stage of drafting the framework, Facebook questioned the EU committee on whether its targeted advertising feature fell into the subliminal system category. This could suggest that the social media giant is aware that it is potentially shaping users’ behaviour, she says.
She notes that most of the EU’s regulations are focused on the middle tier – AI systems that pose a high risk to the public’s health and access to education, as well as the administration of justice and democratic processes.
“If it’s categorised as high risk, you need to have mitigation measures, including being able to show how the data was processed and used to avoid discrimination,” she explains, adding that many tech services will likely fall under this category.
Informed consent – notifying users that their data is being collected and how it will be used – is also a requirement, she says.
Under the proposed framework, requirements for high-risk AI systems cover the quality of data sets used; technical documentation and record keeping; transparency and the provision of information to users; human oversight; and robustness, accuracy and cybersecurity.
In case of a breach, the requirements will allow authorities to access the information to investigate whether the use of the AI system complied with the law.
Closer to home, Singapore is also preparing for AI, releasing the Model AI Governance Framework (Model Framework) for consultation in January 2019 to address ethical and governance issues.
“By explaining how AI systems work, building good data accountability practices, and creating open and transparent communication, the Model Framework aims to promote public understanding and trust in technologies,” states its Personal Data Protection Commission, which is in charge of developing the framework.
“It’s telling that there’s an attempt to get companies to be more accountable, to show that the fears people have about AI can be contained,” says Lau.
She says introducing a framework to manage AI systems is akin to putting a child in a living room – the child must still be confined to a safe space without sharp corners and harmful objects.
“Cost will increase dramatically for businesses to apply the changes that will come with these new laws.
“You want people to keep inventing novel ways to harness technology to make our lives better, but there’s this fear – which is not unfounded – that things could go too far,” she says.
Lau adds that she doesn’t expect technology to reach the singularity point – a hypothetical point when technological growth becomes uncontrollable and irreversible, resulting in unforeseeable changes to human life – anytime soon.
A common example used in science fiction are killer robots like in the Terminator franchise in which the AI SkyNet becomes sentient, views humans as a threat and triggers a nuclear holocaust.
National roadmap
Malaysia too is exploring a framework for AI, originally mooted by the Malaysia Digital Economy Corp (MDEC) in 2019.
Last December, the Science, Technology, and Innovation Ministry awarded an RM400,000 grant to Prof Dr Rose Alinda Alias, a lecturer at Universiti Teknologi Malaysia’s (UTM’s) Azman Hashim International Business School, to formulate the National Artificial Intelligence Roadmap for Malaysia 2021-2025.
The professor led a multidisciplinary team of 20 UTM experts from five faculties along with consultants from the National Tech Association of Malaysia (Pikom) over four months to develop a roadmap that outlines how the country should adopt AI.
Some suggestions put forward include establishing an AI Coordination and Implementation Unit (AI-CIU) that will be responsible for the implementation of the roadmap; adopting current cybersecurity policies and best practices for AI; and establishing a code of ethics.
To protect consumers from AI misuse, Prof Mohammad Ershadul echoes the call for AI-CIU, suggesting that a national taskforce made up of members from various government bodies be set up.
He recommends including the Communications and Multimedia Ministry, Domestic Trade and Consumer Affairs Ministry, and Bank Negara, as it will take a multi-disciplinary unit to manage AI in various spaces such as fintech.
He opines that Malaysian regulators’ conservative approach has been proven sensible, adding that the financial regulators’ stance on cryptocurrency has kept bitcoin scams to a minimum.
From as far back as 2018, Bank Negara has stated that digital assets like cryptocurrencies are not recognised as legal tender, warning that the central bank doesn’t regulate the operations of bitcoin.
It advised the public to be cautious of the risks associated with the usage of such digital currency.
“The average consumer isn’t adequately informed, so the government has to protect them,” explains Prof Mohammad Ershadul.
Self regulating or self serving?
Currently, many tech giants exploring AI have formed committees to self-regulate its use.
Microsoft established its Aether Committee in 2017, and IBM and Google formed their own a year later.
The Aether Committee includes various departments, including the Office of Responsible AI.
This department collaborates with stakeholders within and outside the company to shape new laws, norms and standards to help ensure that the promise of AI technology is realised for the benefit of all.
A paper it released for other companies and its legal advisors, called AI Governance And Ethics For Boards, highlighted that regulation from governments would eventually be a matter of fact, and thus companies needed to be proactive in building trustworthy internal checks.
Prof Mohammad Ershadul warns that self regulation has a tendency to be self-serving too.
By assuring the public that they are responsible, companies look good in the eyes of investors and potentially keep away government intervention, he says.
“We do think that self regulation is admirable but it’s also not remediable. Who will these companies be accountable to when they self-regulate?” Prof Mohammad Ershadul asks.
He sees the future of regulation as a collaborative effort between governments and tech companies, with the process being better served by active dialogues rather than one-sided enforcement.
“I see the future being co-regulation rather than self-regulation. We voted for the government, not for the tech giants, thus co-regulation will put the government in a better position to perform its regulatory responsibilities and have a more informed view of what tech companies are doing,” he says.
He adds that Malaysia will also need to pass additional laws to form a board that could work with tech companies.
He theorises that frameworks for AI regulation would also likely have to be broad rather than black and white, as it is a fast-evolving industry affecting many fields.
“This allows entrepreneurs to enjoy the freedom to develop their goods and services while keeping in mind the boundaries set by the principles.
“Anticipatory or principle-based regulation is both proactive and iterative and is able to respond to evolving markets,” he explains.
However, being too preemptive could stifle innovation, says Lau.
“The idea is to find that balance between innovation, accountability and regulation,” she says.
She also feels AI regulations have to be fluid and ever evolving while avoiding too many loopholes.
“I think it’s going to be something we are constantly going to negotiate. Regulation has to evolve with what it is trying to control,” she sums up.
Already a subscriber? Log in
Get 20% OFF The Star Digital Access
Cancel anytime. Ad-free. Unlimited access with perks.
