Abstract
It is believed that legislation is an inferior technology to technology. Artificial intelligence is shaping our decisions on credit, employment, policing and adjudication in Pakistan. No statute governs it. The dominant view treats this silence as a vacuum that only foreign regulation can fill. This article challenges that view. It argues that the Constitution of Pakistan 1973 already contains the substantive norms needed to govern artificial intelligence, and that what is missing is statutory and institutional machinery. Pakistan’s AI regulation is doctrinally coherent at the constitutional level, doctrinally inadequate at the statutory level, doctrinally incomplete at the data protection level, and doctrinally aspirational at the policy and proposed legislation level. Using a doctrinal-comparative method disciplined by a receivability rule, the article tests the European, Chinese, British and American models element by element. It offers a four-limb statutory definition and a constitutionally filtered hybrid framework for Pakistan.
1. Introduction and Problem Statement
In Pakistan, algorithmic systems are being implemented faster than laws are being developed in reference to them. Public agencies and businesses are beginning to employ automized systems to evaluate, categorize, and forecast. The legal system, which receives these systems, was based on the fact that human being used to be the decision maker. For instance, the Prevention of Electronic Crimes Act 2016 punishes unauthorized access, forgery, and fraudulent actions.[1] Any of these crimes presupposes human source of mistakes. If the system acted in a discriminatory way or misled someone unintentionally, then offense will not be applicable in this case. The amendment of 2025 has contributed to broadened executive control over internet content but has not impacted algorithmic decision-making.[2] The adoption of the Personal Data Protection Bill 2023 will mean that in order to use citizen’s data, the consent of the owner of that data and determination of its purpose will be needed. However, this bill does not imply any regulation similar to Article 22 of GDPR or the right to initiate a human assessment of automated decision-making.[3] The only legislative effort to date is the Regulation of Artificial Intelligence Bill 2024, which fails to define key elements of the AI. This problem has not been clarified in the literature. Comparative scholarships examine each foreign model in its own constitutional context. Domestic writing lists issues of privacy, bias and accountability, but does not develop instruments which can be applied by a court. This is not a problem that has been addressed in the literature. Comparative scholarship is a study of each foreign model within its particular constitutional framework. Domestic writing focuses on issues of privacy, bias and accountability, while failing to develop instruments a court can apply, which is why it is often seen as a failure. The research question is thus specific and precise. What can be made of the constitutional, statutory and treaty obligations that Pakistan already has and has a constitutionally valid basis for doing so?
2. Theoretical Framework and Methodology
The inquiry is doctrinal first and comparative second. It fixes the lex lata before arguing for the lex ferenda, because a reform proposal that skips the existing law is merely a policy preference. The lex lata is read in four layers: constitutional, statutory, data protection and policy. Provisions are construed purposively and read in harmony with Pakistan’s obligations under the International Covenant on Civil and Political Rights, which it ratified in 2010.[4]
The comparative limb rests on a deliberate theoretical choice. The legal transplants debate has settled into two poles. Watson held that rules move easily between legal systems, while Legrand denied that they move at all.[5] Neither position tells a legislator which foreign element to admit. The thesis takes a third position through what it calls the receivability rule. No foreign rule enters Pakistani law unless it survives interpretation against the Pakistani constitutional text. An element passes only if it fits the rule of law in Article 4, the rights in Articles 9, 10A, 14, 19 and 19A, the judicial architecture of Articles 175 and 199, and the State’s real administrative capacity.[6] The test is applied to elements rather than whole models. A provision can thus be received without the constitutional theory that produced it.
3. Core Arguments and Major Findings
The first finding is that Pakistan lacks machinery, not constitutional norms. The thesis maps each harm onto its constitutional anchor. Algorithmic opacity engages the reasoned-decision requirement in Article 10A. Bias engages the equality guarantee in Article 25, read to reach disparate effects without proof of intent. Profiling and surveillance engage Article 14, whose privacy content the Supreme Court developed in Benazir Bhutto through a reasonable-expectation test.[7] The step that carried this test from the home to the telephone now carries it to algorithmic surveillance. In Ishfaq Ahmed v Mushtaq Ahmed the Supreme Court began this work. It warned against opacity, bias and hallucination in judicial use of artificial intelligence and insisted on a human judge.[8] The ruling concerns courts alone. Its reasoning, however, does not depend on who decides, so the thesis extends it to administrative and commercial automated decisions.
The second finding concerns definition. The definition in Article 3(1) of the EU AI Act is so wide that Ruschemeier reads the Act as a law for almost all software.[9] The thesis therefore proposes a four-limb test. An artificial intelligence system is a machine-based system that (a) operates within a defined task domain, (b) learns its decision rules from data through statistical inference, (c) produces outputs that influence its environment, and (d) acts with operational autonomy, so that its output cannot be deduced from its design alone. The test confines regulation to Narrow AI.
The third finding is comparative. No foreign model is receivable in full. The EU risk taxonomy is received as structure. Its product-safety logic is rejected, because it gives affected persons few enforceable rights, a weakness that Smuha and that Veale and Zuiderveen Borgesius have each exposed.[10] China’s iterative technique, algorithm registry and regulatory sandbox are received, while its content-control orientation fails Article 19. The United Kingdom contributes the separation of technical evaluation from rulemaking and five principles that must become binding duties. The USA provides sectoral supervision through the existing regulators; a lesson confirmed by the swift revocation of Executive Order 14110.[11]Where the regulatory object is still in flux, that argument holds true, according to Hagemann, Huddleston Skees and Thierer, who conclude that only soft law can keep up with technology.[12] That argument has force where the regulatory object is still unsettled. No privacy, equality or due process is unsettled. They are legal obligations that must be upheld, not negotiated
The fourth discovery relates to liability. Negligence, strict liability and vicarious liability are already recognised by the Pakistani law. Allocation is the problem, not doctrine. The thesis shifts the burden of strict liability to the deploying entity, maintains a non-excludable recourse against the developer and removes the legal personality of the machine. The victim is thus provided with a local and identifiable defendant.
4. Policy Implications and Conclusion
The results have a different effect on the work of reformers. While a reformer trying to prove the legitimacy of new principles is fighting against a tide, the reformer who provides tools for new ideals that have already been enshrined in the constitution comes from the position of the defender of the current powers. Parliament should therefore enact a dedicated Artificial Intelligence Act rather than amend the 2016 Act or the data protection bill. It should also abandon the 2024 Bill, whose defects are structural. The proposed Act converts each value into a duty with a duty-bearer, a trigger, defined content, a test of adequacy and a consequence in a named forum. Transparency thus becomes a duty to give reasons sufficient to enable challenge. The Act establishes a Pakistan Artificial Intelligence Authority with fixed tenure and ring-fenced funding, and an Artificial Intelligence Tribunal for grave penalties and contested compensation. It expressly preserves writ jurisdiction under Article 199, avoiding the ouster defect in the Digital Nation Pakistan Act 2025.[13] Until Parliament acts, the superior courts can give interim effect to reason-giving and human review, as Ishfaq Ahmed demonstrates.
The research offers three contributions: a definition, a methodology, and a jurisprudential framework. It is limited about the concept that refers to Narrow AI, whose institutions are yet to be created. It draws lessons which far exceed it. A functioning constitutional order is adapted to a novel type of technology via interpretation, not innovation. Rights formulated in 1973 can also govern systems that their authors could not even think of if a legislature establishes the required machinery through which these rights can be exercised.
Writer’s Bio: Rais Nouman Ahmed is an Assistant Professor of Law at Gillani Law College, Bahauddin Zakariya University, Multan. He holds a PhD in artificial intelligence regulation and an LLM from Staffordshire University, UK, and has taught law since 2013. His research covers constitutional law, digital governance and AI regulation. He is regarded as Pakistan’s first domestically trained doctoral specialist in AI regulation. His publications include a Routledge book chapter and HEC-recognised journal articles on AI regulation and on Pakistan’s digital and criminal law.
[1] Prevention of Electronic Crimes Act 2016, ss 3, 4, 14 and 15.
[2] Prevention of Electronic Crimes (Amendment) Act 2025.
[3] Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data (General Data Protection Regulation) [2016] OJ L119/1, art 22; Personal Data Protection Bill 2023 (Pakistan).
[4] International Covenant on Civil and Political Rights (adopted 16 December 1966, entered into force 23 March 1976) 999 UNTS 171, arts 2(3), 17 and 26.
[5] Alan Watson, Legal Transplants: An Approach to Comparative Law (2nd edn, University of Georgia Press 1993); Pierre Legrand, ‘The Impossibility of “Legal Transplants”‘ (1997) 4 Maastricht Journal of European and Comparative Law 111.
[6] Constitution of the Islamic Republic of Pakistan 1973, arts 4, 9, 10A, 14, 19, 19A, 25, 175 and 199.
[7] Mohtarma Benazir Bhutto v President of Pakistan PLD 1998 SC 388.
[8] Ishfaq Ahmed v Mushtaq Ahmed PLD 2025 SC 582.
[9] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) [2024] OJ L 2024/1689, art 3(1); Hannah Ruschemeier, ‘AI as a Challenge for Legal Regulation: The Scope of Application of the Artificial Intelligence Act Proposal’ (2023) 23 ERA Forum 361.
[10] Nathalie A Smuha, ‘From a “Race to AI” to a “Race to AI Regulation”: Regulatory Competition for Artificial Intelligence’ (2021) 13 Law, Innovation and Technology 57; Michael Veale and Frederik Zuiderveen Borgesius, ‘Demystifying the Draft EU Artificial Intelligence Act’ (2021) 22 Computer Law Review International 97.
[11] Executive Order 14110 of 30 October 2023, ‘Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, 88 Fed Reg 75191, revoked by Executive Order 14148 of 20 January 2025, ‘Initial Rescissions of Harmful Executive Orders and Actions’, 90 Fed Reg 8237.
[12] Ryan Hagemann, Jennifer Huddleston Skees and Adam Thierer, ‘Soft Law for Hard Problems: The Governance of Emerging Technologies in an Uncertain Future’ (2018) 17 Colorado Technology Law Journal 37.
[13] Digital Nation Pakistan Act 2025.
