260204P - HUMAN VALUES IN MACHINE MINDS: THE RESEARCHER’S ROLE IN ETHICAL AI DEPLOYMENT TO ENSURE ULTIMATE HUMAN CONTROL

Presented at the 10th International Virtual Medical Research Symposium (MRS) 2026 at the IIUM Kuantan Campus by Professor Omar Hasan Kasule MB ChB (MUK), MPH (Harvard), DrPH (Harvard), Chairman of the Research Ethics Committee at the Natural and Health Sciences Research Center.

Summary

• Human control under the doctrine of taskhiir ensures morality and ethics because the human has superiority over all creation.
• Artificial intelligence and automation, if not well controlled, will threaten human control, resulting in unethical or immoral consequences.
• It is the duty of the scientist who researches artificial intelligence or uses AI in other areas of research to make sure that human moral control prevails.
• We have to guard against the misuse of AI for immoral purposes.

Triad of khilafat, taskhiir, and umran - 1

• The start of human history was the elevation of Adam to be khalifah, a responsibility that all of his descendant inherited until today. This elevation was based on moral superiority and superior knowledge as told in the Qur’an, Baqara: 31-37. 

• وَإِذْ قَالَ رَبُّكَ لِلْمَلَائِكَةِ إِنِّي جَاعِلٌ فِي الْأَرْضِ خَلِيفَةً ۖ قَالُوا أَتَجْعَلُ فِيهَا مَن يُفْسِدُ فِيهَا وَيَسْفِكُ الدِّمَاءَ وَنَحْنُ نُسَبِّحُ بِحَمْدِكَ وَنُقَدِّسُ لَكَ ۖ قَالَ إِنِّي أَعْلَمُ مَا لَا تَعْلَمُونَ (30) وَعَلَّمَ آدَمَ الْأَسْمَاءَ كُلَّهَا ثُمَّ عَرَضَهُمْ عَلَى الْمَلَائِكَةِ فَقَالَ أَنبِئُونِي بِأَسْمَاءِ هَٰؤُلَاءِ إِن كُنتُمْ صَادِقِينَ (31) قَالُوا سُبْحَانَكَ لَا عِلْمَ لَنَا إِلَّا مَا عَلَّمْتَنَا ۖ إِنَّكَ أَنتَ الْعَلِيمُ الْحَكِيمُ (32) قَالَ يَا آدَمُ أَنبِئْهُم بِأَسْمَائِهِمْ ۖ فَلَمَّا أَنبَأَهُم بِأَسْمَائِهِمْ قَالَ أَلَمْ أَقُل لَّكُمْ إِنِّي أَعْلَمُ غَيْبَ السَّمَاوَاتِ وَالْأَرْضِ وَأَعْلَمُ مَا تُبْدُونَ وَمَا كُنتُمْ تَكْتُمُونَ (33) وَإِذْ قُلْنَا لِلْمَلَائِكَةِ اسْجُدُوا لِآدَمَ فَسَجَدُوا إِلَّا إِبْلِيسَ أَبَىٰ وَاسْتَكْبَرَ وَكَانَ مِنَ الْكَافِرِينَ (34) وَقُلْنَا يَا آدَمُ اسْكُنْ أَنتَ وَزَوْجُكَ الْجَنَّةَ وَكُلَا مِنْهَا رَغَدًا حَيْثُ شِئْتُمَا وَلَا تَقْرَبَا هَٰذِهِ الشَّجَرَةَ فَتَكُونَا مِنَ الظَّالِمِينَ (35) فَأَزَلَّهُمَا الشَّيْطَانُ عَنْهَا فَأَخْرَجَهُمَا مِمَّا كَانَا فِيهِ ۖ وَقُلْنَا اهْبِطُوا بَعْضُكُمْ لِبَعْضٍ عَدُوٌّ ۖ وَلَكُمْ فِي الْأَرْضِ مُسْتَقَرٌّ وَمَتَاعٌ إِلَىٰ حِينٍ (36) فَتَلَقَّىٰ آدَمُ مِن رَّبِّهِ كَلِمَاتٍ فَتَابَ عَلَيْهِ ۚ إِنَّهُ هُوَ التَّوَّابُ الرَّحِيمُ (37)

• The angels had doubts about the moral superiority of humans. This is because humans can make mistakes because of their free will; however, they can correct them by atonement, unlike angels, which are automatons and have no free will. 

Triad of khilafat, taskhiir, and umran - 2

• The knowledge superiority was shown by teaching Adam the names of all things which angels did not know. 
• The knowledge superiority of humans was directly related to taskhiir, the ability of the human to control and modify his environment using the large brain capacity and special anatomical features that enable making. 
• Taskhiir enables humans to build a physical civilisation (‘imarat al ardh or ‘umran). Human civilization has the potential of being moral because morality is encompassed within khilafat.

Taskhiir

• The term taskhiir is used in the Qur’an 27 times. It is very comprehensive. It implies that Allah put all creations, biological and physical, at the service of humans. The creations are endowed with properties that facilitate their control by humans and make them at the disposal of humans. 
• The heavens and earth are under human control. [Qur'an, Luqman: 20].

أَلَمْ تَرَوْا أَنَّ اللَّهَ سَخَّرَ لَكُمْ مَا فِي السَّمَاوَاتِ وَمَا فِي الأَرْضِ وَأَسْبَغَ عَلَيْكُمْ نِعَمَهُ ظَاهِرَةً وَبَاطِنَةً. [لقمان: 20]

• Oceans and vessels are under human control. [Qur'an, Al Jaathiyat: 12].

•  {اللَّهُ الَّذِي سَخَّرَ لَكُمُ الْبَحْرَ لِتَجْرِيَ الْفُلْكُ فِيهِ بِأَمْرِهِ وَلِتَبْتَغُوا مِن فَضْلِهِ} [الجاثية: 12].

• Animals under human control. [Qur'an, Hajj: 36].

• وَالْبُدْنَ جَعَلْنَاهَا لَكُم مِّن شَعَائِرِ اللَّهِ لَكُمْ فِيهَا خَيْرٌ... كَذَٰلِكَ سَخَّرْنَاهَا لَكُمْ لَعَلَّكُمْ تَشْكُرُونَ} [الحج: 36 {سُبْحَانَ الَّذِي سَخَّرَ لَنَا هَـٰذَا وَمَا كُنَّا لَهُ مُقْرِنِينَ. [الزخرف: 13]. 

Levels of autonomy in AI

1. Fundamental questions to ask: Is AI a tool or is it intelligence? Does it have a personality?  Does it have autonomy?
2. Autonomy broadly means the ability of AI systems to operate without human control and supervision. There are 3 levels of autonomy, each with different ethical implications: "human-in-the-loop" (co-pilots), "human-on-the-loop" (supervisory), and "human-out-of-the-loop" (fully autonomous). 
3. Human in the loop and human on the loop are forms of automation that we are familiar with, manifesting as automation but within a script set by and controllable by the human at will. 
4. Both human-in-the loop and human-on-the-loop are considered partial autonomy. Partially autonomous AI can analyse data, make real-time decisions, take actions, etc.
5. "Human-out-of-the-loop" (fully autonomous) is a theoretical possibility with serious consequences because it violates taskhiir. There is no empirical evidence of fully autonomous AI. 

Human control (taskhiir) vs AI autonomy

• If humans control technology, we are assured that ethical values will be respected. The recent developments in AI have questioned human control (taskhiir by the Qur’an).
• Technology is achieving autonomy in decision-making, with machines able to manufacture other machines that produce goods or carry out procedures without human intervention. 
• AI can analyze large data sets and reach conclusions without taking moral values into consideration, such as future vision and benefits, which require human wisdom (hikmat). 
• Technology may end up reversing taskhiir and controlling the human. In that eventuality, who then will take responsibility for mistakes? We will require ethical guidance to answer these questions.
• As long as AI operates within limits set by ethical humans, we are safe. Loss of human control can be due to neglect, and as soon as the human wakes up to responsibilities, control can be reasserted. 

Causes of loss of human control in IA

• Loss of human control may be due to moral decay, such that the human is no longer able to control even if he wants because of a lack of moral criteria to use in judging safety, harm, and benefit.
• Loss of control may be due to interference by shaitan, who has limited and temporary power over humans when their iman is weak or absent. Shaitan can mislead anybody except the committed ones. [Qur'an, Sad: 82-83].

• قَالَ فَبِعِزَّتِكَ لَأُغْوِيَنَّهُمْ أَجْمَعِينَ إِلَّا عِبَادَكَ مِنْهُمُ الْمُخْلَصِينَ. [صاد ٨٢-٨].

• Widespread sinning decreases iman and makes the misuse of AI easier. Sins are committed in the absence of eeman. [Bukahri 6782].

•  لَا يَزْنِي الزَّانِي حِينَ يَزْنِي وَهو مُؤْمِنٌ، وَلَا يَسْرِقُ السَّارِقُ حِينَ يَسْرِقُ وَهو مُؤْمِنٌ.  [البخازي ٦٧٨٢].

Ethical issues of partially autonomous AI with easy human control

1. The ethical problems of AI are more likely due to human moral transgressions. A prime example is deepfakes and other forms of misrepresentation.
2. Partially autonomous AI can be directed by humans towards doing good, but can also be directed to do bad. In this case, AI is a tool in human hands.
3. Selfish use of AI for commercial profit, which is very clear in advertising AI as the solution to all problems.
4. AI has become a fad. As we know, every few years fads come as fashionable and disappear.

Ethical issues of partially autonomous AI with difficult human control

1. Human control is affected by the volume of data involved because the human mind can store and process a finite amount of data, while AI’s data memory is very extensive. The human may not follow what AI is doing to be able to control it.
2. Human control is challenged by the speed of processing, which is higher in AI, such that the human cannot keep up with control.
3. When the human loses control, shaitan in the form of humans can take control and mislead individuals and communities

Fully autonomous AI with impossible human control

• Human out of the loop (fully autonomous) is a theoretical concept and is the major issue in discussing autonomy vs taskhiir. If it can occur, it violates taskhiir and also violates the moral foundations of khilafat and umraan, resulting in potential destruction of the cosmic order.
• Fully autonomous AI will raise legal and ethical issues of mistakes without attributing them to a human operator, being a complete denial of taskhiir.

Sources of ethical values about AI

1. The researcher must observe two ethical values that are relevant today: purposiveness (ghaiyat) and avoiding unquestioning blind following (taqlid). 
2. Ethical values are from religion. The technological part of AI does not require input from transmitted knowledge ('ilm 'naqli); it is purely human reason (ilm aqli). The contributions of ilm naqli is in underlying guiding concepts and ethics that will guide the technology and stop it from causing grievous harm. 
3. Muslims have the responsibility to disseminate these concepts to the rest of humanity to push back against harm to human civilization. 
4. How can we incorporate ethics in AI algorithms?

Lack of a future vision

• Each discovery leads to another automatically, and every innovation is accepted without questioning; nobody wants to be left behind. 
• There is no overall vision of where we are going. We had no vision 10 years ago of what we see today. 
• Is AI more intelligent than humans?

Blind following

1. Technology has become an automaton that cannot be stopped if it generates revenue for the owners. The Qur’anic term for this situation is ‘abath (26:128, 23:115). 
2. Blind following of everything new in technology before asking serious moral questions is condemned today as it was condemned by the Qur’an in previous centuries (2:170-171, 33:67). 
3. No critical and selective absorption of technology.

Qaidat al dharar: harms of AI as a tool - 1

• End-users of AI products and processes must be informed of the use of AI, and its potential mistakes and adverse effects, and must give their informed consent
• Misleading info with the help of AI. Harm of deepfakes.
• AI and crime: cause and prevention
• Injustice when AI is used to harm some groups and benefit other groups
• AI makes the commission of crime easier and more efficient, but can also help prevent crime.
• AI can be misused for social evil.

Qaidat al dharar: harms of AI as a tool - 2

• Taking human control out of the equation raises the issue of criminal liability for mistakes made by AI.
• AI makes the protection of privacy and confidentiality more difficult. Security vs privacy.
• AI is used to produce more lethal weapons.
• Artificial lover/mother/father.
• Mass unemployment /enslavement of human laborers (Sulaiman and the jinn).
• Practical results that are not explicable. Where did they come from? No transparency, explainability. Black box nature of algorithms.

Qaidat al dharar: harms of AI as a tool - 3

• AI will ignore the dignity or humanity of the patient.
• Inclusiveness of the evidence.
• Speed and efficiency.
• Errors.
• AI as a learning/teaching tool.
• AI and qualitative research.