Editor's note: Rostam J. Neuwirth is a professor of the Faculty of Law, University of Macau. The article reflects the author's opinions, and not necessarily the views of CGTN.

The 1945 Constitution of the United Nations Educational, Scientific and Cultural Organization (UNESCO) declares that "since wars begin in the minds of men, it is in the minds of men that the defences of peace must be constructed." Presently, the same kind of thinking should be applied to one of the most serious global challenges that humanity currently faces, which is the future regulation of artificial intelligence (AI). In November 2021, UNESCO took a first important step and adopted the Recommendation on the Ethics of Artificial Intelligence as a first global standard-setting instrument on AI, which recognizes "the profound and dynamic positive and negative impacts of AI on societies, environment, ecosystems and human lives, including the human mind."

The emphasis on the impact of AI on the human mind is of utmost importance, because – in line with advances in neuroscience – there already exist many AI techniques in a great number of areas of the lives of billions of people with profound consequences, which, however, most of the time are not noticed. Worse still, many of these techniques are intentionally designed to "subliminally" – that is below the threshold of awareness – manipulate the human mind and to alter a person's behavior.

Such subliminal techniques are commonly known from experiments carried out in the 1950s in the field of advertising, which used rapidly flashed images in movie theaters to boost the sales for popcorn and Coca-Cola. The book "The Hidden Persuaders," published in 1957, warned that the same subliminal techniques can be used not only to boost the sale of products but also to influence the outcomes of elections. By the late 1980s, several jurisdictions around the world banned the use of "subliminal advertisements" in audio-visual media. More recent scientific experiments not only confirmed the effectiveness of subliminal techniques and their general impact on the human mind and behavior, but also highlighted the possibilities for their application in a wider range of areas.

For instance, a software called "brain spyware" illustrates how it is possible to access private information, such as banking information, PIN codes, location of residency or the date of birth, from the brains of brain–computer interface users "by showing them visual stimuli and using a machine learning model to detect familiar information based on the brain's response." Equally, so-called dark patterns describe undetectable deceptive elements used in websites and apps to manipulate users through various methods, such as disguised ads, triggering fear or "bait and switch." Or else, the combination of targeted advertising with hidden algorithms used by search engines was found to determine not only the success of business ventures but also the outcomes of elections, as it was revealed by the Cambridge Analytica Scandal.

In brief, these are but a few examples and there already exist many more sophisticated techniques capable of manipulating the human mind. AI and notably future AI systems will enhance these possibilities further. This will be made possible by the creative combination and convergence of different technologies, tools and applications directly and indirectly related to AI, such as machine learning, affective computing and eye-tracking.

For these reasons and to secure the consistency with the ban of subliminal advertising, it is therefore necessary to also prohibit subliminal techniques of AI in a future global standard-setting instrument on AI. Failing this, it should at least be duly considered in every legal system of the world. As a first important precedent, the European Union Proposal for an Artificial Intelligence Act already contains a model clause, which specifically aims to prohibit the "use of an AI system that deploys subliminal techniques beyond a person's consciousness in order to materially distort a person's behavior in a manner that causes or is likely to cause that person or another person physical or psychological harm."

The sophistication of technologies able to both "subliminally" and "supraliminally" manipulate the human mind has already reached unprecedented levels. The rapid pace of innovation in the technologies underlying AI are going to further increase the impact of AI on our lives in general and the human mind in particular. In other words, what today may still sound like the story of a science fiction novel will – as history repeatedly showed – be reality tomorrow, such as the use of AI as a "digital crystal ball" allowing their developers to predict the future of everyone by manipulating their behavior.

For these and more reasons, it is urgent to elaborate on a legally binding global standard setting instrument on AI, which addresses various manipulative practices in general and particularly the concrete questions of how to define "subliminal AI systems," where to draw the lines between desirable and harmful forms of manipulation, how to balance individual vis-a-vis societal harm, and, most importantly, for what purpose future AI shall be developed. Not attending to these issues bears the existential risk of human intelligence becoming replaced by AI. It would also allow to subliminally manipulate the minds of the masses in violation of the right to freedom of thought aimed to protect the privacy of the mind and human thoughts.

(If you want to contribute and have specific expertise, please contact us at opinions@cgtn.com.)