Skip to content
Screenshot 2026-09-08 at 20.42.10

Pakistan

Tools Tools
AI Attorney | ANPR | ChatGPT | CFMS | Claude | Digilawyer | Digital Wakeel | Face Trace | Facial recognition technology | Gemini | Judge-GPT | LRC-Assistant | NAZ Assist | Next Gen Safety App | Pakistan Law Bot | Punjab Emergency AI | QanoonAI | Specialist systems | Safe City | SkyEye | Your Munshi
Tasks Tasks
Administrative support | Data review and analysis | Legal research, analysis, and drafting support | Operational support | Predictive analytics
Law enforcement | Courts | Victims Users
Law enforcement
Scope Scope
Nationwide
Training Training
No mandatory or systemic training
Regulation Regulation
No dedicated legislation governs AI in criminal proceedings. The 2025 National AI Policy and April 2026 Judicial AI Guidelines address AI use generally but not criminal justice specifically. Potentially relevant evidentiary provisions exist under the Qanun-e-Shahadat Order 1984 (Articles 164, 46-A)
Insight Insights
Islamabad's use of AI-powered SkyEye surveillance system has been credited by the police with approximately 2,700 AI-linked arrests and a nearly 45% improvement in police dispatch rates since 2023
Cases Cases
The Supreme Court has addressed AI's role in adjudication in Ishfaq Ahmad v Mushtaq Ahmad (PLD 2025 SC 582), cautioning against bias risks in AI-assisted bail/sentencing tools and affirming that AI cannot substitute judicial decision-making, while encouraging development of clear guidelines. In Abdul Salam Khan v Bank Al-Habib Ltd (PLD 2025 SC 1043), it reiterated that efficient case management using AI tools is a constitutional responsibility. Lower courts have informally used AI (e.g. ChatGPT) to assist legal research, including in a 2023 Punjab injunction and juvenile bail matter
Information uploaded as at July 2026

AT A GLANCE

Pakistan is gradually incorporating AI into criminal proceedings, with the most developed use case found in law enforcement, where Safe City projects in Lahore, Karachi, Islamabad, and elsewhere in Punjab combine CCTV, facial recognition, and predictive hotspot analysis through centralised command-and-control hubs. Judicial use is more limited but growing, with AI-powered legal research tools (including JudgeGPT and the Sindh High Court's search engine) assisting judges, and the Chief Justice having announced a strategic initiative to integrate AI further into the judiciary. There are no reported instances of prosecutors or defence counsel using AI, although informal use of AI research tools is likely. AI use by victims is limited to administrative and legal-aid chatbots.

Pakistan does not yet have AI-specific legislation governing criminal proceedings. The 2025 National AI Policy outlines a general developmental and regulatory agenda but does not address criminal justice specifically, and the NationalGuidelines for Use of Artificial Intelligence in Judicial Institutions of Pakistan (2026) set out permissible judicial uses of AI. The Islamabad AI Declaration (2026) is a policy document setting out Pakistan’s core principles and execution priorities for deployment of AI in digital governance, making clear references to AI subservience to Constitutional objectives and of human oversight in matters of public consequence involving the use of AI. In the interim, AI-generated or AI-processed evidence may be brought within general evidence law under the Qanun-e-Shahadat Order 1984. Pakistan has not formally engaged with the UNESCO Guidelines for the Use of AI Systems in Courts and Tribunals (2025), though the Judicial Guidelines state that they align with international best practice.

USE

As at July 2026, AI use and institutional development in Pakistan’s criminal justice system is most developed in law enforcement, driven by ‘Safe City’ projects in Lahore, Karachi, Islamabad, and elsewhere in Punjab, which combine CCTV footage, facial recognition technology, and predictive hotspot analysis through centralised command-and-control hubs. The deployment of AI has reportedly had a measurable impact in crime reduction and effective response time of the police.

While the deployment and subsequent impacts of AI in law enforcement are more known and reported, its deployment in the judicial realm is limited and narrowly confined. The judiciary, prosecutors, defence counsel, and victim-facing services show more limited, largely discretionary or pilot-stage AI adoption.

Law enforcement

The use of AI by law enforcement agencies in Pakistan remains in its nascent state with the focus currently on real-time visual identification and spatial crime mapping (hotspot analysis), and limited evidence of individualised risk assessment mechanisms.

Predictive analytics

Predictive analytics tools for law enforcement agencies have been deployed in a select few cities.

In the Federal Capital of Pakistan, the SkyEye system—an AI-driven digital monitoring and facial recognition system—has been deployed to transform Islamabad’s Safe City Project. Since 2023, the system has reportedly led to an almost 45% improvement in police dispatch time, and has resulted in approximately 2,700 AI-linked arrests. The system has reportedly also led to a reduction in both heinous and property crimes. Islamabad Police have also launched the ‘Safe City Hunch Lab’, which processes emergency and call data and then examines information based on location, time, and type of crime to pinpoint hotspots and ascertain crime patterns.

Similar predictive analytics tools have been deployed in Punjab Safe City zones, particularly Lahore, through Punjab Emergency AI, a system linked to pre-existing Safe City infrastructure and used to identify crime hotspots and zones where crime is likely to be high, allowing for efficient police patrol plans. The system is designed to integrate three years of First Information Report (police complaints report “FIRs”) data, safe city camera feeds and real-time feed intelligence to enable more efficient and effective allocation of resources.

Data review, analysis and operational support

The Safe City Project, operating in collaboration with Huawei, has led to the installation of surveillance cameras across key urban centres including Lahore, Karachi, Islamabad, and other cities in Punjab. These cameras are strategically placed to cover key entry and exit points, commercial hubs, government sites and residential zones. AI-enabled systems process data acquired through the cameras and provide operational support in several key phases:

  • Data review: a ‘Talk to Camera feature allows police officers to interact with the entire camera network to identify issues such as suspicious activity or health hazards such as open manholes or unusual smoke.
  • Recognition and detection: the system is integrated with facial recognition technology and automatic number plate recognition technology, which can be used to identify and capture the movement of individuals and vehicles. The system operates with the support of the National Database and Registration Authority.
  • Data restructuring: CCTV feeds are processed by AI to provide geo-tagging and event-tagging.
  • Output: data is integrated into Pakistan’s Integrated Command, Control and Communication (‘IC3’) centers, which are subsequently used by law enforcement and emergency response services.

The IC3 centres provide a centralised outlet for various security and rescue services to act on real-time data. Recent developments by the Federal Investigation Agency claim to allow authorities to ascertain the visual appearance of known figures in real time, using AI-based tools to generate likely appearances.

The SkyEye system in Islamabad (discussed above) allows officers to reconstruct suspect movements across dispersed camera networks, allowing law enforcement to trace individuals through sequential visual data points. Arrangements with the National Database and Registration Authority also allow law enforcement to match stored facial images against their own database.

The Punjab Information and Technology Board has introduced Face Trace, which is an AI-based facial recognition system for the Punjab Police. It provides real-time matching of a person with various datasets. The data compiled for the system includes 16 million records and pictures from the driving licenses branch of government, 1.8 million records from the government’s criminal records database, 1.3 million records from the Punjab Khidmat Markaz (police facilitation centre), and 300,000 records of accused individuals and criminals from Punjab prisons.

AI is also being used for surveillance and monitoring purposes in criminal investigations, expediting working timelines to identify suspects. In Islamabad, the police reported that through the use of AI, the police force were able to make arrests within 24 hours of a murder. The investigation involved the use of geofencing (a location-based technology creating a virtual boundary around a physical geographical area), analysing footage from 250 cameras, and reviewing 130 phone call records, all reportedly supported through the use of AI.

Contemplative Gaze with Futuristic Colors

Prosecutors

As at July 2026, there have been no reported uses of AI by prosecutors in Pakistan. However, at the 23rd Pakistan Prosecution Forum, prosecution leaders highlighted efforts to modernise the criminal justice system through AI-enabled case management and prosecution information systems, which will assist prosecutors in identifying causes of case delays, provide workload management rolls to distribute cases more efficiently, and integrate data systems to improve coordination across justice institutions.

Courts

AI use in courts in Pakistan is, as at July 2026, limited. However, the Chief Justice of Pakistan’s Supreme Court, Mr. Yahya Afridi, has announced a strategic initiative to integrate AI into the country’s judicial system to enhance efficiency, transparency, and service delivery.

Legal research, analysis and drafting support

There is a growing number of legal research, analysis and drafting support tools that help judges and lawyers harness the potential of AI to augment their efforts in researching case laws and understanding procedural pathways.

LRC-Assistant (an AI-enabled Search Engine) and CFMS (Mobile App on iOS Platform) were launched by the Sindh High Court to modernise judicial administration and ensure access to justice for all. This development marked the first case of a High Court introducing an in-house, AI-powered legal search engine and decision support system. It leverages trained machine learning models and indexed judicial data repositories through an intelligent chatbot interface to help access orders, case diaries and judicial records. Similarly, CFMS is a litigant and lawyer-friendly iOS platform that provides lawyers and litigants with easy access to court services and information.

JudgeGPT is a custom generative AI assistant designed for Pakistan’s trial courts by Dr. Sultan Mehmood (a professor at the New Economic School) in collaboration with Professor Elliot Ash and researchers at ETH Zurich working alongside Pakistan’s Federal Judicial Academy. It is a chatbot based on OpenAI’s GPT-4 family of models that assists judges with legal research and drafting. The tool was deployed on a test basis for use by 1,559 judges serving across 118 courts, representing the first large-scale field experiment evaluating the integration of generative AI into a national justice system. The study introduced judges to the generative AI assistant by dividing them into one of three categories : (i) access to the AI assistant with targeted training tailored to its use; (ii) access to AI with generic training on technology and law; and (iii) generic training without access to AI. At median-district exposure, introducing AI with targeted training corresponded to 1,848 additional cases resolved per year, a 6.3 % increase over the mean.

Another tool available for judges and lawyers is QanoonAI, an AI-powered legal platform with a reported high percent citation accuracy. It combines a database of more than 186,000 Pakistani judgments with 43 AI-powered tools and features a free legal guidance chatbot in Urdu.

Pakistan Law Bot is an AI-powered legal platform and research assistant built specifically for Pakistani law and for use by Pakistani courts and lawyers. It provides tools for automated legal research, document drafting and case management, and connects users with a verified market place of local advocates. It features a chatbot that allows users to ask questions in natural language and receive guidance with citations to Pakistani laws and judgments. The tool is bilingual, operating in both English and Urdu.

AI Attorney is an AI-powered platform integrated with ChatGPT and Claude that provides assistance in researching Pakistani case law, drafting court pleadings, reviewing documents and organising cases with citation-aware legal intelligence. Other legal research tools in Pakistan that integrate AI functionalities and which may be used by judges include Digilawyer, Digital Wakeel, and Your Munshi may also be used by lawyers and judges in Pakistan.

Defence

Legal research, analysis and drafting support

Though no official uses of AI by defence counsel in Pakistan have been reported as at July 2026, it is becoming common for defence counsels to consult commercial AI systems including those identified above (see ‘Courts’ section), to gain assistance in legal research, drafting and to determine a general sense of direction with regards to the potential development of a matter as it moves through the court system. In particular, Your Munshi is an AI legal assistant that provides assistance in drafting, monitoring and court date management. There may also be informal use of commercial AI systems such as ChatGPT, Gemini, and Claude.

Victims

Administrative support

NAZ Assist is an AI tool developed by Pakistan’s Legal Aid Society to support citizens with free, 24/7 legal advice. The Society claims it operates within an ecosystem of their own in-house programmes including gender-based violence, women’s right to legal property, and protecting and promoting religious freedoms.

The Safe City Project (discussed above) has introduced a Next Gen Safety App, integrated within the Safe City Project’s AI-powered surveillance system to provide access to rescue and emergency services, a panic button, location sharing and also access to free legal advice. The Safe City system has also integrated an AI chatbot for citizen queries regarding criminal records, legal advice, and available police services.

The Sindh High Court launched CFMS, a user-friendly iOS platform for court users with iPhones, which permits lawyers and their clients to retrieve case information and file documents electronically, which is meant to support digital access to justice through enhancing transparency and reducing delays.

TRAINING

As at July 2026, there is no mandatory or systematic training on the use of AI by actors in the Pakistani criminal justice system. The National Guidelines for Use of Artificial Intelligence in Judicial Institutions of Pakistan (discussed below) mention training of judges in the use of AI, although they are not a statute conferring a binding necessity of training of judges in the use of generative AI. There have, however, been ad hoc initiatives.

For example, for judges, the Punjab Judicial Academy has conducted training programmes addressing ‘Generative AI for Legal Decision Making’. Moreover, as discussed above, judges that consented to a one-year subscription to JudgeGPT received training on the use of the tool. The National Judicial Policy Making Committee has announced, as part of its new guidelines (discussed below), that training will be provided to judges as part of the justice system’s digitalisation and capacity-building within judicial institutions.

For lawyers, private universities have offered training sessions on the use of AI in practice. For example, the Punjab Information Technology Board has launched an ‘AI-Based Litigation Management & Training System’ for the Law & Parliamentary Affairs Department. Training sessions were also offered to various public sector legal departments, including District Attorney’s offices. Additionally, the Legal Aid Society has partnered with EnablifyAI to provide independent practitioners with training seminars to help them use AI and eventually develop their own AI database to improve their respective practices and streamline their workflow. Furthermore, EnablifyAI, in collaboration with Kilam Law, a leading Legal 500 ranked law firm, has also focused on AI enablement for in-house legal teams owing to the burgeoning realisation that for in-house teams, legal expertise alone is not enough and that for more efficient and active provision of legal support to clients, the use of AI in developing and managing workflows has become an indispensable reality.

For law enforcement, UNODC Pakistan organised in September 2024 a workshop on the use of AI to combat terrorism aimed at building the capacity of National Counter Terrorism Authority analysts. Furthermore, under-training Assistant Superintendents of Police (ASPs) were duly trained in AI-based policing as part of their vocational training with Punjab Safe Cities Authority which concluded on 21st March, 2025. Additionally, the Ministry of the Interior of Pakistan has announced a partnership with Beijing’s Police Department to train members of the Islamabad police on how to use AI to support their policing work.

WhatsApp Image 2025-10-02 at 16.18.36

REGULATION

As at July 2026, there is no specific legislation governing the use of AI in Pakistan. However, some policies and guidelines provide the basis for development of a specific legal framework for the use and deployment of AI in the broader judicial and governance structures of the country.

Islamabad AI Declaration (2026)

The Islamabad AI Declaration represents an important development in the area of AI use and deployment, as it represents the official intention of the State to transition from high-level policy goals into structured national execution. The Declaration enumerates nine pillars that are to guide the advancement of digital society, digital economy and digital governance. The first three pillars are relevant to the justice sector:

  1. Sovereign Purpose, Public Value – Under this pillar, AI shall serve Constitutional purposes while being utilized for strengthening national productivity, improving governance and delivering measurable public outcomes.
  2. Human Accountability, Constitutional Authority – AI is to augment human capacity, not replace lawful authority. Decisions of Public consequence shall remain under accountable human oversight, supported by review, due process and institutional responsibility.
  3. Pragmatic Delivery, Use Case First – AI shall be deployed where it demonstrably improves economic productivity, service delivery, operational execution and decision support. Scale is to follow proof of impact.

The Declaration establishes the foundations for AI governance and supervision in Pakistan and reflects a disciplined commitment to sovereignty, public trust, and measurable national value. Pakistan will adopt AI responsibly, govern it rigorously, and build domestic capability with accountability.

Dr. Sohail Munir, Chairperson of the Pakistan Digital Authority, February 2026

Screenshot 2025-10-03 at 13.23.30

Guidelines for practitioners

As at July 2026, though guidelines have been issued for judges (discussed below), to our knowledge there are no guidelines issued by local, provincial, or national Bar Councils in Pakistan regarding the use of AI by lawyers. In the absence of specific guidance, practitioners are bound by the Legal Practitioners and Bar Councils Act 1973, and the professional conduct rules thereunder, as well as general ethical obligations their respective Bar Councils apply to them. Under Section 41 of the Legal Practitioners and Bar Councils Act 1973, an advocate may be reprimanded, suspended, removed from practice, fined or penalized if found guilty of misconduct. The Act further empowers the Pakistan Bar Council to make rules under Section 55, pursuant thereto are the Pakistan Legal Practitioners and Bar Council Rules 1976. Rule 161 provides:

An advocate shall not intentionally misquote to a judge, judicial officer or jury the testimony of a witness, the argument of the opposing advocate or the contents of a document; nor shall he intentionally misquote to a judge or judicial officer the language of a book, statute or decision; nor shall he, with knowledge of its invalidity and without disclosing such knowledge, cite as authority a decision that has been over-ruled or a statute that has been repealed or declared unconstitutional.

The above provision may be attracted in situations where the use of AI has led to hallucinated citations or an advocate relying on AI has overstated an argument or misconstrued a matter of the record. Further, Rule 163 mandates that an advocate engaged in public prosecution has a primary duty to “see justice is done”, not pursue a conviction or suppress or conceal facts. Rule 175-A confirms that a breach of any of the rules relating to conduct including Rules 161 and 163 qualifies as misconduct under the parent Act.

National Guidelines for Use of Artificial Intelligence in Judicial Institutions of Pakistan (2026)

In April 2026, the National Judicial Policy Making Committee published the National Guidelines for Use of Artificial Intelligence in Judicial Institutions of Pakistan. In drafting the Guidelines, the National Judicial Policy Making Committee was led by Justice Muhammad Ali Mazhar of the Supreme Court of Pakistan. The Guidelines set down the permissible uses of AI in the Judiciary and key guiding principles for judicial use, and are applicable to all superior courts, lower courts, special courts, tribunals and judicial institutions. The guiding principles include:

Human oversight

AI is to be used as a tool to augment judicial functions and not to replace the decision-making of a judge. All processes involving AI are to include human checkpoints where the judge is to give independent reasoning so as to not allow AI to replace critical judgment.

Ethical integrity and transparency

AI systems used in judicial processes are to be transparent and free from algorithmic and data bias. Users in the judiciary must be able to identify, with references, the specific portions of the document that have been generated by AI and explain the information verification process. To ensure this, a standardized system will be used to ensure traceability, accountability and avoiding plagiarism.

Privacy and data security

Use of AI is to be strictly in compliance with all relevant data protection and privacy laws wherever applicable. Confidential information of litigants is to be secured against unauthorized access, use or disclosure.

Accountability and oversight

Judges, court staff and other users are to be imparted training to understand the capability and limitation of AI tools.

Permissible uses of AI by judges include:

Administrative functions

Case management, scheduling, automated document routing, translation purposes, transcribing minutes of meetings, and generating administrative reports.

Legal research

Al-powered legal search engines and tools that assist judges in finding relevant case law, statutes, and legal commentaries.

Predictive analytics

AI tools for predicting case durations or identifying case patterns to assist in resource allocation and case management.

Judicial assistance

AI systems that provide initial summaries of case documents, draft sample legal texts, and assist in organising evidence.

The Guidelines mention preference for locally developed systems trained on Pakistani legal data.

The Guidelines are non-exhaustive. They do not explicitly prohibit any other use of AI by judges, however, any other use must comply with the Guidelines’ overarching safeguards. In particular, AI must not replace a judge’s final decision-making authority or substitute generative output for independent judicial reasoning

The National Judicial Policy Making Committee has stated that it respects the administrative and judicial autonomy of High Courts, allowing each jurisdiction to tailor implementation according to its needs.

As at July2026, Pakistan does not appear to have formally engaged with the UNESCO Guidelines for the Use of AI Systems in Courts and Tribunals (2025) in an official capacity. However, the National Judicial Policy Making Committee's guidelines expressly state that the framework ‘aligns with international best practices’. Moreover, UNESCO welcomed the Guidelines and stressed that they ‘draw upon UNESCO’s Recommendation on the Ethics of Artificial Intelligence’.

Screenshot 2026-08-22 at 08.20.36

Criminal procedure rules

The use of AI in criminal proceedings in Pakistan may be governed by: (i) the Code of Criminal Procedure (1898), which is the procedural law covering criminal trials; and (ii) the Qanun-e-Shahadat Order (1984), the law governing evidence.

The Code of Criminal Procedure (1898)

As at July 2026, there appear to be no explicit provisions in the Code of Criminal Procedure 1898 regarding the admissibility of AI-related evidence in Pakistan. There are however several provisions regarding the use and admissibility of evidence during criminal justice proceedings which may be relevant. Section 94 of the CrPC permits a court or officer in charge of a police station to require the production of any document or other thing considered necessary or desirable for an investigation, inquiry or trial.

Section 96 empowers the Court to issue a search warrant where it believes that the person to whom a summons or order under Section 94 will not produce the document or other thing as required.

The Court is further empowered under Section 104 to impound any document or other thing produced. Section 244(2) allows a Magistrate, on application of the complainant or the accused, to issue summons to any witness directing him to attend or to produce any document or other thing.

While the term ‘thing’ or ‘other thing’ has not been defined within the CrPC, the Lahore High Court in Shakeel v The State etc (PLD 2025 Lahore 430) has given the following guidance as to its interpretation at Paragraph [34]: ‘In particular, the term things used in section 94 CrPC. and Article 161 of QSO must be given a broad meaning and understood to signify anything connected with the offence whose production will serve the interest of justice.’

Qanun-e-Shahadat Order (1984) (the law of evidence)

The Qanun-e-Shahadat Order (1984) is concerned with the admissibility and weight of evidence in Pakistan generally. As at July 2026, the Order does not contain any provisions that explicitly deal with AI related evidence. Two provisions are, however, relevant and could possibly be utilised to bring AI-generated or AI-processed evidence before a court.

  • Article 164 gives the court a general discretion to permit the production of evidence that has become available as a result of modern devices or techniques, provided the court considers it appropriate to do so in the circumstances of the case.
  • Article 46-A separately establishes that statements taking the form of electronic documents are relevant facts where they were generated, received, or recorded by an automated information system that was functioning properly at the time.

Taken together, these two provisions could plausibly be relied upon to bring AI-generated or AI-processed evidence before a court.

With respect to the status of evidence received or stored electronically in an automated information system, the following provisions provide for its weight and admissibility.

  • Article 73 in its Explanation Clauses 3 & 4 provides that a printout or any other form of output of an automated information system shall not be denied the status of primary evidence solely for the reason that it was generated, sent, received or stored in electronic form if the automated information system was in working order at all material times. Any such material shall be treated as primary evidence where a security procedure was applied to it.
  • Article 78-A provides that if an electronic document is alleged to have been generated wholly or in part by any person through the use of an information system, and where such allegation is denied, the application of a security procedure to the signature or the electronic document must be proved.

The term ‘security procedure' is defined in the Electronic Transactions Ordinance 2002 in Section 2(x) as a procedure agreed between the parties, implemented in the normal course of business for establishing the authenticity and integrity of any electronic document. This definition is endorsed in the Qanun-e-Shahadat Order.

Pakistan Penal Code (Act XLV of 1860)

In February 2025, Punjab police registered three FIRs under PECA arising from the alleged uploading of AI-generated deepfake videos and images depicting Punjab Chief Minister Maryam Nawaz, charged additionally under Sections 500, 504, and 505-1(C) of the Pakistan Penal Code.

In the above context, Section 500 delineates the punishment for defamation as being simple imprisonment which may extend to 2 years and/or a fine. Similar penal consequences are attracted to print, engrave, or sell matters known to be defamatory in Sections 501-502.

Section 504 relates to intentional insult which provokes or intends to provoke any person to breach public peace.

Section 505-1(c) criminalizes the publication or circulation of a statement, rumour or report with the intent to incite any class or community of persons.

Whilst the Pakistan Penal Code does not explicitly provide for AI related offenses or the use of AI in criminal proceedings, the following provisions may also be attracted:

  • Section 191 deals with the provision of false evidence where a person is legally bound to state the truth, for example in a court of law.
  • Section 192 criminalises the fabrication of false evidence.
  • Section 203 provides for imprisonment up to 2 years and/or a fine where an accused provides false information with respect to the matter they have been accused of.

Data protection legislation

The existing data protection legislation in Pakistan is centred around cybercrime and digital governance as opposed to any direct framework regulating automated decision-making, automated data collection or the protection of an individual’s rights in situations of surveillance. Various statutes and policies outline the regulatory steps that will be taken, but it remains to be seen how they will complement the existing data protection and human rights frameworks with respect to AI systems.

Future reforms may arise. Pakistan’s National AI Policy 2025 (discussed below) details, in Chapter 3, the Government’s plans to implement human oversight mechanisms for critical AI operations, conduct compliance audits and define a legal framework with specified penalties and incentives. It also details plans to tackle generative AI use and a national data security policy.

Data Protection Bill (2023)

Pakistan’s Data Protection Bill was first drafted in 2023 by the Ministry of Information Technology and Telecommunication. As at July 2026, as per media reports, the Bill remains unenacted and subject to government revision and stakeholder consultation. Though the Bill does not contain provisions specific to the use of AI in criminal proceedings, if enacted, the Bill will address issues of privacy and consent regarding data processing, sharing, and storage.

  • Section 2(j) defines a data subject as a natural person who is the subject of personal data.
  • Section 2(z) defines personal data as any information that relates directly or indirectly to a data subject, who is identified or identifiable from that information or other information in the possession of a data controller and/or data processor, including any sensitive or critical personal data. It is further clarified that anonymized or pseudonymized data incapable of identifying a person is not personal data.
  • Section 2(kk) classifies biometric and genetic information, health information, identification documents, criminal records, religious beliefs, political affiliations and castes, tribes and ethnicity information as sensitive personal data.
  • Section 2(g) defines critical personal data as personal data retained by the public service provider - excluding data open to the public - as well as data identified by sector regulators and classified as critical or any data related to international obligations.
  • Sections 5-9 of the Bill outline the consent required from the relevant person, the relevant notice requirements and impose obligations upon the data controller to protect the personal data.
  • Section 14 imposes the requirements to process a child’s personal data in a manner that protects their rights and interests and includes requirements for consent of the child’s parents or guardians.
  • Section 15 regulates the use of sensitive personal data and critical personal data. It shall not be processed unless the data subject has given explicit consent and any of the further requirements of Section 15(a) are met.
  • Sections 16-29 establish the withdrawal of consent, prevention of processing where it may cause damage and distress, the right to request erasure of data and the grievance redressal process. Section 29(4) preserves the right not to be subject to a decision based solely on automated processing which, if enacted, may be a useful safeguard against AI-related profiling and data collection.

Digital Nation Pakistan Act (2025)

The Digital Nation Pakistan Act (2025) is positioned as an enabling statute to facilitate a shift towards the development of digital infrastructure. It envisions a Digital Authority (Sections 6 and 7), developed to oversee and regulate data management, governance, and AI. However, the details regarding the Digital Authority, including the scope of its AI oversight mandate and its enforcement mechanisms, are limited.

  • Section 8(g) restricts the Digital Authority to ensure that there is no overlap with the jurisdiction of other regulatory bodies and no conflict with relevant laws or existing regulatory frameworks.
  • Section 12 enables the authority to monitor implementation of the Digital Nation Pakistan Act and ensure compliance through formal notices specifying corrective action to public entities. However, no penalties or offences have been outlined under the Act or any delegated legislation as of yet.
  • Section 28(2) confirms that existing laws regarding data protection and cybersecurity shall hold precedent.

As at July 2026, the Pakistan Digital Authority has been established. Whilst it has not as of yet provided for any rules or regulations, a draft National Data Governance Policy 2026 is under consideration. The draft policy assigns the Pakistan Digital Authority with responsibility for issuing binding standards, conducting compliance audits and directing corrective action.

Cybersecurity laws

Prevention of Electronic Crimes Act (2016)

The Prevention of Electronic Crimes Act 2016 (PECA) is the primary cyber security statute in Pakistan. It deals with cybercrime through the criminalisation of: identity theft and misappropriation of identity information in Section 16, dishonest misuse of information through Sections 3 - 8, and crimes relating to human dignity and sexual offences in Sections 20 – 24. Section 26-A criminalises the intentional dissemination, public exhibition or transmission of information which the person knows, or has reason to believe, is false or fake and which is likely to cause fear, panic, disorder or unrest among the public or society. The statute further provides for the establishment of an investigation agency (the NCCIA) and elaborates on the agency’s powers to investigate (Section 30 & 30D), provisions for victim and witness protection (Section 30B) as well as broader powers regarding remand (Section 30A), warrants (Section 33-34) and information handling (Section 36, 39).

Although PECA does not expressly regulate AI, conduct performed through or facilitated by AI may fall within the Act where all the constituent elements of a particular PECA offence are established. The use of AI does not, by itself, render conduct criminal.

Human rights

The Constitution of Pakistan (1973) serves as the primary human rights framework in the country by enshrining certain fundamental rights. While there is no mention of AI in the Constitution, nor any amendment addressing AI specifically, several provisions may still be relevant. Most significantly, Article 14 protects the right to dignity and privacy, through which, the Constitution guarantees the right to privacy and dignity which may serve as a safeguard against AI-generated non-consensual explicit consent and overreach in predictive policing. Article 9 (protecting the right to life and liberty), Article 19 (freedom of speech), Article 19A (right to information), and Article 10A (the right to a fair trial) may also be invoked.

Article 25A provides that all citizens are equal before the law, and further prohibits discrimination on the basis of sex (Article 25A(2)). Article 25 may be engaged where an AI system used by a public authority produces unjustifiably different outcomes for similarly situated persons, including through biased training data or proxy characteristics. Context-specific discrimination safeguards also appear in Article 26 (Non-Discrimination in respect to public access) and Article 27 (Safeguards against discrimination in services)

Additionally, Pakistan has binding obligations as a State Party to the International Covenant on Civil and Political Rights (ICCPR), where Article 2(1) requires each party state to respect Covenant Rights without distinction on grounds including race, colour, sex, language, religion, political or other opinion, national or social origin, property, birth or other status.

  • Article 26 of the ICCPR protects equality before the law, equal protection of the law and protection against discrimination.
  • Article 17 protects against arbitrary or unlawful interference with privacy, family, home or correspondence and against unlawful attacks on honour and reputation, making it relevant to biometric surveillance and data profiling.

Moreover, under the UN Convention on the Rights of the Child:

  • The rights and interests of children are broadly protected under Articles 2 and 3.
  • Article 16 protects children against arbitrary or unlawful interference with privacy, family, home or correspondence and against unlawful attacks upon honour and reputation.
  • Article 40 provides more specific protections for children accused or recognised as having infringed the criminal law, including treatment consistent with their dignity, the presumption of innocence, fair determination of the allegations and respect for their privacy at all stages of the proceedings. These provisions may be engaged by facial recognition, biometric databases, automated profiling or the use of a child’s personal information in an AI system.

Code and Bokeh Lights-1

Outlook

In 2025, the Ministry of Information Technology & Telecommunication presented its National AI Policy which represents a pivotal milestone for transforming Pakistan into a knowledge-based economy. The Policy provides a developmental framework that ‘coherently interlaces with the draft Personal Data Protection Act’. The Policy envisages a secure ecosystem ‘to ensure AI serves the public good, freedom of information, protects individual rights, and aligns with national priorities.’

The Policy does not address AI use in criminal proceedings specifically. Its only reference to the justice sector appears under the Fourth Pillar (‘Transformation and Evolution’), which lists ‘Governance’ as one of several priority areas for sectoral AI adoption roadmaps, encompassing policing, equal access to justice and information, and ‘quick decision making’ among civil servants — without further elaboration, dedicated safeguards, or an implementation timeline specific to criminal justice. However, general safeguards proposed under the Secure AI Ecosystem pillar — including human oversight for high-risk applications, a public register of government AI systems, lifecycle impact assessments, and third-party algorithmic audits — would presumably apply to any future justice-sector deployment.

CASES

Possibilities for AI adoption

In Ishfaq Ahmad v Mushtaq Ahmad (PLD 2025 SC 582), the Supreme Court of Pakistan explored the potential benefits and concerns surrounding the use of AI in the judicial process whilst also stressing the need for a guiding framework. The following observations of the Supreme Court provide insight into AI adoption and constitutional compliance:

  • ‘Under Articles 10A and 37(d) of the Constitution of the Islamic Republic of Pakistan, 1973 (“Constitution”), the right to fair and expeditious justice must not be rendered illusory. Within this constitutional framework, the thoughtful adoption of AI can serve as a viable instrument for access to timely justice and alleviating systemic backlogs.’ (para. 7)
  • ‘AI, when deployed within principled boundaries, holds significant potential to enhance judicial and institutional productivity and efficiency.’ (para. 9)
  • ‘The right to a fair trial before a competent, independent, and impartial judge is a fundamental principle of due process. AI must not overshadow the core guarantee of judicial autonomy. While AI has the potential to improve consistency and efficiency in legal processes, it also carries the risk of introducing biases and limiting judicial discretion.’ (para. 12)
  • It was unequivocally stated that AI can never become a substitute for judicial decision-making, by affirming that ‘AI tools remain subordinate to judicial reasoning and must never be mistaken as a substitute for the exercise of judicial conscience, discretion, or interpretive judgment. It is strongly underlined that the judicial role requires that AI tools be viewed as supplementary aids rather than substitutes for human reasoning. The responsibility for ensuring the accuracy, ethical integrity, and confidentiality of judicial determinations rests entirely with the judge.’ (para. 9) It also noted that: [i]t must be unequivocally affirmed that AI is not, and must never become, a substitute for judicial decision-making. Adjudication is a constitutional function rooted in legal reasoning, institutional independence, and human empathy, qualities no automated system can replicate. The human part of justice lies in its moral and emotional core, the qualities that AI can never replicate (at least today).’ (para. 13)

The Supreme Court encouraged the use of AI tools in court proceedings and recommended that the National Judicial (Policy Making) Committee in collaboration with the Law and Justice Commission of Pakistan develop clear and principled guidelines for courts and practitioners to follow. The Court stressed that the guidelines ‘must delineate clear boundaries, ensuring that AI is used only as a facilitative tool and never in a manner that compromises human judicial autonomy, constitutional fidelity, or public trust in the justice system.’ (para. 18)

In Abdul Salam Khan v Bank Al-Habib Ltd PLD 2025 SC 1043, the Supreme Court again briefly spoke on how the delays in adjudication are a key challenge in judicial governance. The Court at Paragraph [7] held that a ‘modern, responsive, and intelligent case management framework ’ is a constitutional responsibility of the judiciary. In the same paragraph, it reiterated the need for judicial use of AI tools to assist in scheduling and triage.

Prior to the Supreme Court’s jurisprudence on this matter, there have been earlier documented instances of AI use in the lower courts of Pakistan. For instance, in 2023, an Additional District and Sessions Judge in the Mandi Bahaudin District used ChatGPT to explore grounds for granting an injunction in Civil Appeal No. 11 of 2023. While his decision was not based on his use of AI, he did express support for implementing AI as an assisting technology in the judicial process.

AI-generated evidence

In Mir Hassan v The State (2025 YLR 645), the Balochistan High Court refused to place reliance on video evidence presented by the prosecution on the basis that the prosecution had not sent it for forensic analysis to determine if it was doctored or an outcome of AI. This reflects the growing caution that judges approach cases within criminal matters owing to the apprehensions of doctored evidence as a result of AI use.