THE Federation of Malaysian Consumers Associations appreciates the perspective of the science, technology, and innovation minister in his article “Navigating the Future: Malaysia’s Ethical Vision”. He recognises the growing impact of AI on society and the challenge of balancing the pursuit of innovation with the protection of human values and wellbeing.

I would like to build on these issues and further elaborate on the potential harms of AI on society at large. These include:

Opaque systems and lack of accountability

Currently, these systems are highly opaque, and consumers have no information on how the system works or the basis of decision-making. The general lack of transparency could have significant effects on consumers.

Mistakes and inaccurate data

Generative models do not “understand” context; thus, the content they produce may look convincing and correct but be factually incorrect. It can be difficult for the person using or promoting the content to notice these errors if they are not already familiar with the facts of the relevant subject.

Deepfakes and disinformation

As generative AI models become increasingly powerful, it becomes easier to use them to create realistic synthetic images, text, or voice recordings that can be mistaken for real content. In early 2024, a Hong Kong company employee transferred USUS$25 million to a scam account based on instructions from a deepfake image of the Chief Executive Officer. A 2022 European report estimates that by 2026, about 90% of online content may be AI-generated. As the volume of synthetic content grows, it becomes difficult to trust one’s own eyes and ears. The long-term effects of this can be devastating on trust in institutions and each other.

AI in advertising

The adverse effects of AI on consumers include making it easier and more efficient to manipulate people through personalised advertising.

Bias and discrimination

AI models can perpetuate or create new biases. AI models are trained on vast amounts of information from the internet, thus they inherit the biases of their training data. As such, the models will generate content that reproduces bias, negative, or unwanted tendencies.

Privacy and data protection

Personal data has long been coveted as highly valuable for businesses to use in targeted advertising to individuals and groups. When generative AI models are trained on material scraped from the internet, the training data usually contains large amounts of personal data. As generative AI models are developed and deployed, the issues related to data protection and personal data can lead to substantial privacy harms.

Security vulnerabilities and fraud

Generative AI models can be abused by malicious actors to augment or supercharge criminal activities. Generative AI models can be used to make fraud, scams, and other illegal activities more efficient.

Environmental impact

Tech companies are already emitting a substantial amount of carbon. AI companies use a substantial amount of energy and water. It has been reported that training a single model can consume more electricity than 100 US households use in an entire year. This exponential use of energy and water has a severe negative impact on the environment. Clearly, AI technology comes with a high carbon footprint.

Further, the minister mentioned three regulatory framework models: the rights-based model, the state-driven model, and the market-based model. However, at this point in time, Malaysia will be following a fourth form of regulatory framework – an ethics-based framework. Generally, that would mean a set of “ethical guidelines” and tech companies need to voluntarily adhere to those guidelines. This approach assumes that tech companies have consumers’ and society’s wellbeing as a core component of their growth and development. Is it true – do tech companies really have society’s wellbeing at heart?

Past experiences in other jurisdictions suggest otherwise. While the European Union was the first region to develop laws to regulate AI, there have been reports of intense lobbying by tech companies with immense resources to water down any form of mandatory requirements. Their rationale would be that regulation would weaken “innovation” and their preferred regulation is self-regulation. Even in the US, there has been intense lobbying to weaken any form of regulation that would impose mandatory requirements on the industry.

Technology is not an untameable beast but must be adapted and shaped by the rules and values ensuring consumer protection and consumer wellbeing. To ensure that generative AI is developed and used in accordance with consumer and human rights, it is clearly insufficient to rely on companies to regulate themselves. It is the responsibility of policymakers and enforcement agencies to set boundaries for how technology is trained, developed, deployed, and used. Therefore, policymakers must pass laws and regulations which are necessary to provide safe and consumer-centric technology in the years to come.

To ensure that generative AI is safe, trustworthy, fair, equitable, and accountable, there is a need for overarching principles that address consumer rights. The principles set below provide a foundation for how policymakers and enforcement agencies should approach the opportunities and pitfalls of generative AI.

These rights can be summarised as follows:

• Consumer rights must be respected.

• Consumers must have the right to object and to an explanation.

• Consumers must have the “right to be forgotten” to have personal data deleted.

• Consumers must have the right to interact with a human instead of generative AI.

• Consumers must have a right to redress and compensation for any damages.

• Consumers must have the right to collective redress.

• Consumers must have the right to complain to supervisory authorities or launch legal actions.

Developers and deployers of generative AI models must establish systems to ensure that these rights are available.

Regulation of AI is still at an early stage of development. However, some proposals to protect consumers and society from other jurisdictions include:

Mandatory disclosure

Companies deploying generative AI should be required to disclose prominently when AI is used in generating content or recommendations. This transparency allows consumers to make informed decisions and understand the source of the information they are consuming.

Quality and accuracy standards

Establish quality and accuracy standards for content generated by AI systems. Companies should be held responsible for ensuring that the content generated by their AI systems meets these standards. This could include measures to minimise misinformation, bias, or harmful content.

Liability for harmful content

Define liability for companies whose generative AI systems produce harmful or misleading content. This could include provisions holding companies accountable for damages caused by content generated by their AI systems, particularly in cases of defamation, infringement of intellectual property rights, or dissemination of false information.

Data protection and privacy

Strengthen data protection and privacy provisions to safeguard consumer data used to train generative AI systems. Companies should be required to obtain explicit consent from users before using their data for training AI models and ensure that data privacy rights are respected throughout the process.

Monitoring and enforcement

Allocate resources for monitoring the use of generative AI systems by companies and enforcing compliance with the regulations. This could involve establishing specialised units within regulatory agencies tasked with overseeing AI-related issues and conducting regular audits of companies’ AI systems.

Consumer education and awareness

Launch public awareness campaigns to educate consumers about the capabilities and risks of generative AI technology. Empowering consumers with knowledge about how AI works and its potential impacts can help them make more informed choices and protect themselves from potential harm.

We call on policymakers and lawmakers to take a strong stance in favour of consumer protection and preserving human rights. It is necessary to have robust legal measures, including strict obligations on developers and deployers of generative AI systems, to operate in a transparent and accountable manner and to restrict the development, deployment, and use of systems that are fundamentally incompatible with these rights.

Clearly, in the battle between “innovation” and “protection of human rights”, tech companies with their deep pockets would intensively lobby for “innovation” and weak or no regulation.

The minister has rightly indicated that the journey to shape an AI-enhanced world that respects human dignity and promotes societal wellbeing cannot be undertaken by the ministry alone but requires a collective effort by all stakeholders.

Recognising the extreme power of big tech and their focus on “innovation” and the need for “weak regulation” to facilitate “innovation”, we call on the minister to invest in creating a society wellbeing task force and build its capacity to be able to realistically negotiate with big tech to protect human values and society’s wellbeing.

Due to the rapid and intense evolution of AI, the task force would keep the ministry updated on:

• Latest AI developments, specifically the potential for harm on society.

• Learning how other jurisdictions are regulating AI and providing policy advice to Malaysian regulators

• Coordination with other institutions, for example, Asean or Asia Pacific region, on possible collective actions.

It cannot be denied that without building strong local capacity, the interests of powerful AI tech companies will overwhelm the interests of society at large. A strong consumer voice will, to some extent, help to bring about a stronger voice for society’s wellbeing in that balance. – May 30, 2024.

* Paul Selva Raj is Fomca deputy president.