Pakistan - National AI Policy (2025)

Overview

The National Artificial Intelligence Policy 2025 establishes Pakistan’s first national strategy for AI following Federal Cabinet approval on 30–31 July 2025. Anchored on six strategic pillars, the policy aims to create an inclusive AI innovation ecosystem that balances rapid adoption with safety, human rights and national security. Key structural measures include a National AI Fund (NAIF) to finance R&D and commercialization; a national AI Council and implementation directorate to manage the Master Plan and Action Matrix; geographically distributed Centres of Excellence for compute, training and incubation; regulatory sandboxes for supervised testing; and sectoral roadmaps focused on health, education, agriculture and governance. Official material and government summaries can be found on the Ministry of IT & Telecom website and the Press Information Department. For the MoITT announcement see Ministry of IT & Telecom – AI Policy 2025 announcement.

Definitions

The policy provides operational definitions to ensure common understanding across government, industry and academia. Core definitions cover "Artificial Intelligence" (systems that display human‑like cognitive functions such as learning, reasoning or perception), "AI system" (software or models performing tasks with varying levels of autonomy), "developer/provider" (entities that design, develop or maintain AI systems), "user" (actors who deploy or operate systems), and "high‑risk application" (AI used in health diagnostics, critical infrastructure, social services, law enforcement, financial decisions, or other contexts that materially affect rights or safety). These definitions align with international practice while allowing Pakistan‑specific adaptation in implementation guidance.

Governance and Institutional Framework

Implementation is centred on an AI Council (ministerial‑level, chaired by the Federal IT Minister) and a dedicated implementation directorate to operationalize the Master Plan and Action Matrix. The AI Council will coordinate across federal ministries, provincial governments, regulators (including telecommunications and financial regulators), academia, civil society and industry. The policy mandates creation of Centres of Excellence (CoE‑AI) in major cities to consolidate compute resources, host incubators, and provide training and research facilities. The NAIF is proposed as ring‑fenced financing (official summaries reference a dedicated allocation drawn from existing R&D funds managed by Ignite); separate venture and innovation funds will catalyse private investment. The MoITT is identified as the lead agency for national coordination, with participating roles for HEC, provincial education authorities, Ignite and sectoral regulators. For official ministry details see MoITT announcement and the government press office briefing at Press Information Department – MoITT / Meta event.

Key Focus Areas

The policy’s six pillars (AI Innovation Ecosystem; Awareness & Readiness; Secure & Trustworthy AI; Transformation & Evolution; AI Infrastructure; and International Partnerships) translate into priority programs. Innovation measures: NAIF, CoE‑AI network, incubation, commercialization pathways and IP support. Skills and inclusion: national training programs (target: 1M trained by 2030), train‑the‑trainer initiatives, 3,000 scholarships annually, internships and industry placements. Secure AI & risk: regulatory sandboxes, AI safety guidance, model and data security protocols, and alignment with data protection frameworks. Sectoral transformation: pilot and scale programs in health (tele‑diagnostics, decision support), education (personalized learning), agriculture (yield forecasting, supply chain) and governance (service automation, revenue administration). Infrastructure: national compute grid, centralized and curated datasets, and support for a national LLM to ensure cultural & linguistic relevance. International engagement: bilateral partnerships, participation in global AI standard forums and alignment with multilateral norms. Media summaries and independent analysis provide more detail on targets and pillars; see coverage at Dawn – Cabinet approval and analysis at Business Recorder.

Implementation Framework

Operationalizing the policy relies on three instruments: (1) the Master Plan and Action Matrix (timebound activities, KPIs and responsible agencies); (2) ring‑fenced financing through NAIF and associated venture vehicles; and (3) institutional capacity building (AI Council, implementation directorate, CoE‑AI network). The policy envisages close coordination with HEC, provincial education bodies for curricula alignment, Ignite for R&D financing, and sectoral regulators for conformity and safety standards. Regulatory sandboxes are time‑limited, supervised environments to test models before broader deployment. Procurement rules for public sector AI adoption will require risk assessment, transparency and human oversight clauses. The MoITT page and press briefings describe the governance architecture and next steps; see MoITT announcement.

Monitoring and Evaluation

An Action Matrix with KPIs, regular progress reporting to the AI Council, independent evaluation of pilots, and public dashboards are proposed as the monitoring backbone. The policy mandates annual performance reviews, mid‑term evaluations and third‑party audits of funded projects. Public reporting aims to increase transparency on use of funds, outcomes of Centres of Excellence, and progress towards training and product targets. The AI Council will publish consolidated progress reports and coordinate inspection or impact assessment work with sectoral regulators and, where relevant, the parliamentary oversight committees.

Penalties, Liability, and Appeals

The National AI Policy itself is a high‑level administrative roadmap and therefore primarily prescribes governance, conformity pathways and oversight mechanisms rather than a detailed criminal penalty schedule. It directs implementing agencies and regulators to adopt enforcement measures, administrative sanctions and liability frameworks proportionate to harm and aligned with existing laws (e.g., cybercrime, sectoral regulation and the pending Personal Data Protection framework). The policy calls for development of conformity assessment regimes and redress mechanisms, including consumer complaint channels, remediation obligations for developers/providers, and referral to appropriate judicial or regulatory enforcement under existing statutes. Specific fines, criminal sanctions or precise penalty amounts are to be defined by subsequent regulations and sectoral rule‑making under the authority of relevant regulators.

Relationship to Other Instruments

The policy explicitly situates itself alongside existing legal instruments and institutional frameworks. It directs alignment with Pakistan’s data protection initiatives (the Personal Data Protection Bill / draft law and the National Commission for Personal Data Protection as they evolve), the Prevention of Electronic Crimes Act (PECA) for cyber offences, sectoral rules for health/finance, and relevant international standards. The policy instructs agencies to ensure AI deployment is consistent with constitutional rights and existing labour, consumer protection and information security laws. Government communications emphasise that the policy is a coordinating roadmap — implementing regulations will clarify overlap and precedence with other instruments; see government summaries at PID and MoITT release at MoITT.

International Alignment

The policy expressly seeks alignment with international best practices and standards (e.g., ITU AI for Good, OECD AI principles and other multilateral instrument guidance) and proposes active participation in global fora. It encourages bilateral R&D partnerships, cross‑border research projects and harmonization efforts so Pakistani AI development is interoperable with international markets. The policy also frames the NAIF and venture instruments to attract diaspora and foreign investment under applicable laws, subject to national security and data protection considerations.

Implementation Timeline

Sources and References