AI and Human Rights in Pakistan: Algorithmic Accountability through a
Rights-Centred legal Framework

Artificial intelligence now plays a significant role in the governance, service delivery and
business decision making process in Pakistan.[1] While AI can be efficient and innovative, there
are potential risks to privacy, equality, due process and freedom from discrimination that come
from its unchecked use.[2] The government policies of Pakistan already acknowledge the
importance of transparency, auditability and human oversight, but Pakistan still does not have a comprehensive rights-based approach for algorithmic accountability.[3]
It is argued in this article that Pakistan should adopt a rights based legal framework for
regulating AI, based on the principle of constitutionalism, international human rights norms and
mandate for feasible institutional protections.[4] It suggests a structure that includes impact
assessments, explainability, independent audits, human review, data governance and remedies that are easily accessible for those who are impacted.[5] Artificial Intelligence, Human Rights, Algorithmic Accountability, Pakistan, Privacy, Equality,
Due process, Transparency.In Pakistan, the use of AI has moved beyond being just a theoretical possibility; it is a reality
that is transforming various sectors of society.AI is not just a future reality in Pakistan; it’s a
present one that’s reshaping the industry. [1] The National AI Policy integrates AI into the
national agenda, and puts it in the context of ethical use, fairness, transparency, accountability,
privacy, security, and human rights. The Islamabad AI Declaration also asserts that AI must
enhance human abilities, that decisions with impacts on the public must be made by human
accountable decision makers and that public sector systems that impact rights must be
explainable and auditable, and proportionate to risk.[2] AI systems can impact decisions in the
areas of employment, welfare, policing, education, access to services, etc., but often not be easily discerned in terms of how outputs are produced. This lack of transparency presents important concerns of equality and dignity, privacy, and procedural fairness.[3]
The article addresses the human rights concerns arising from the use of AI in Pakistan and
suggests a framework for algorithmic accountability in Pakistan that meets the rule of law’s
demands.[4][5] The risks of AI to human rights When AI is trained on data that reflects and perpetuates social inequities or it is built poorly, AI
poses a risk to equality. The focus on algorithmic decision-making in Pakistan has raised
concerns about the potential for algorithms to exacerbate exclusion and inequality if not properly safeguarded.[3] Another key issue of concern is privacy. There is a tendency for AI systems to be very data intensive, behavior-driven and monitored.[1] But the government has taken a proactive step in addressing the issue of privacy and security as objectives of governance in the National AI Policy.[4][5] The issue of due process is related to an automated or semi-automated decision that impacts a person without him or her being aware of the reason or how to dispute the decision.[1] The Islamabad AI Declaration tackles these issues by calling for human oversight, review and due process for decisions with public consequences, and incomplete legal processes are in place to ensure the application of these principles.[2] Policy and Legal gaps in Pakistan Pakistan’s policy approach is shifting towards a more rights-centric approach to governance of
AI. Fairness, transparency, accountability, data protection, privacy, security and respect for
human rights and rule of law are all mentioned in the 2017 National AI Policy. It also includes
principles of transparency and human oversight in high-risk situations, lifecycle assessments,
impact assessments, periodic audits, and a program of auditing the decision-making process of AI and redress.[1] The Islamabad AI Declaration further reinforces this position, by asserting that government AI must be explainable, auditable, risk-proportionate, and accountable to the human, and that it must be overseen by human actors. It is also an indication of how the state is responsible for coordinated governance and for sovereign data stewardship, both of which are applicable for fostering institutional responsibility for AI.[2] But civil society has called for more strict response to the law. Both the Artificial Intelligence Bill and the Digital Nation Pakistan Act 2025 need to be substantially amended, Amnesty International has argued, calling for a review of both to ensure they are in line with international human rights law. It is an intervention to highlight the difference between policy words and rights protection.[6] However, Pakistan has yet to have a comprehensive legal framework which explicitly governs the use of high-risk AI. The National AI Policy is not a complete legislation; it contains no claims of liability, remedies, sanctions, or a delineation of institutional competence, among other things; it is more of a policy statement. Towards a Rights Based Framework The government of Pakistan is required to introduce a special law on Artificial Intelligence (AI) that will be risk-based, and assign accountability to those who use AI.[4] The following types of usage should be considered high risk: use in public sector decision making; law enforcement; welfare allocation; credit scoring; employment screening; and any system with the potential of having a material effect on rights or opportunities. Such systems shall have more strict legal obligations than consumer systems that pose a low risk.[5] Five fundamental obligations should be imposed in the framework. First of all, deployers should do human rights impact assessments prior to deployment. First, they should communicate about the use of AI in the transparent and readable manner.[2] Thirdly, decisions of major consequence should be explained, and subject to review by a human decision maker. Third, systems need to be auditable, having logs, test records and bias reports to be kept for access. Fifth, there should be a mechanism for complainants/remedies for those affected.[6] Good data governance should be governed by a strong framework. Certainly policy language already refers to the concepts of sovereign stewardship, privacy, security and trusted exchange of data. These principles need to be translated into obligations to minimise data, guide data use, ensure data accuracy, manage retention, provide security measures, and more special protection for sensitive data. There should be further restrictions on the reuse of data where it is not done with the consent of the data provider or where there is no clear legal basis.[4][5] Algorithmic accountability must have support from institutions. The government of Pakistan should create a separate body on artificial intelligence or designate one of its existing regulators to have the powers to inspect, audit, investigate the complaints, make binding decisions and make guidelines for compliance.[2] The concept of a third-party auditing, public registers and lifecycle assessments outlined in the National AI Policy offer a useful guide for such an institution.[1] The oversight body should work with data protection, telecom, procurement and regulators of the sector.[2] If the state obtains AI tools from vendors, a procurement contract should specify the sharing of responsibility for the errors or harms resulting from the use of the tool, cooperation in conducting technical disclosure, and cooperation in conducting performance testing. This is reinforced by the Islamabad AI Declaration which has a focus on coordinated governance.[4] Only in serious cases should judicial review be available. Courts must be capable of assessing the legality, proportionality, procedural fairness and reason for the decision made using AI assistance. [3]These are especially important when a person is unable to assess themselves the basis for the decision due to opacity of the underlying system or proprietary nature of the system.Effective governance and regulation of AI is nearing its transition from policy to law in Pakistan.
[1]The moment is fast approaching when AI governance and regulation will be transitioning
from policy to law in Pakistan. In the National AI Policy and the Islamabad AI Declaration, there
is clear desire to place human oversight, transparency, auditability and privacy protections on the AI systems, but these need a statutory framework to enable them to defend rights on the
ground.[2] A rights-based legal framework for algorithmic accountability should thus include some form of risk-based regulation, implementing impact assessments, making algorithms explainable,vinstituting independent audits, public registers, human review and effective remedies.[4] Such a system, if practiced by Pakistan, will allow for the promotion of innovative applications of AI, while maintaining the principles of constitutionalism, equality, privacy, dignity, and due process.[5]

References :

[1] Amnesty International commentary on the Artificial Intelligence Bill and the Digital Nation Pakistan Act 2025.

[2] Ministry of Information Technology and Telecommunication, Islamabad AI Declaration (9 February 2026) accessed 8 July 2026.

[3] Ministry of Information Technology and Telecommunication, National Artificial Intelligence Policy (Pakistan) accessed 8 July 2026.

[4] OECD, AI Principles (2019) accessed 8 July 2026.

[5]Algorithmic accountability in Pakistan: constitutional challenges to automated governance (PDF article).

[6]UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021) accessed 8 July 2026.

Writer:

Injila Abid LL.M Researcher, BZU Multan, Punjab, Pakistan. Session: 2025-2027