AI governance begin?

Over the past few weeks, social media in Pakistan has exploded with frustration. Drivers across Lahore and Karachi have been posting screenshots of unexpected challans, debating wrongful fines, sharing blurry images captured by traffic cameras and asking the same question, "How did I get this ticket?" Some report receiving multiple challans at once, while others argue the system is unfair, confusing or technologically flawed.

Albeit underneath this noise lies something far more important. Pakistan has just launched its first real test case of AI-driven governance? The public's reaction exposes deep anxieties about automation, fairness and trust in public institutions. This is not merely a story about traffic challans; it marks the beginning of a new time, one where algorithms make the first decision in enforcement, and where citizens become visible through data. And people of Pakistan are not ready for what this means.

The camera-based e-Challan systems in Lahore, under the Punjab Safe Cities Authority, and in Karachi, under the new Faceless E-Challan System, rely on Automatic Number Plate Recognition (ANPR). This is an AI-powered technology that detects pre-defined violations, reads number plates, cross-checks registries and issues challans without any human intervention.

This is country's first AI bureaucrat and it does three things that the old system could not:

One, it eradicates discretion, thus reducing the potential for corruption.

Two, it increases enforcement coverage, since cameras catch what traffic police officers usually miss.

Three, and most significantly, it creates digital evidence that previously did not exist.

Though this sounds like a success, and in many ways, it is. However, this swift shift toward non-human enforcement discloses Pakistan's weakest link, that our governance infrastructure is not yet fully ready for automated decision-making.

When an AI-based system issues fines, people expect perfection. But ANPR accuracy depends heavily on lighting, non-damaged number plates, camera angles, updated vehicle registries and well-matched digital records. A single mismatch can trigger a domino effect, and the wrong person may be fined, the wrong address may be used, or the wrong evidence may be attached – something that we are seeing periodically on social media.

This is exactly what has happened. The outrage on social media is not only about the fines themselves, rather it is about being governed by a vague system. Citizens feel they cannot appeal, cannot understand the process, and cannot be heard. This is where Pakistan's AI governance gap becomes unmistakably clear.

Countries with advanced AI governance frameworks, for instance the UAE, China, Singapore and Malaysia, build such systems on three pillars:

First is transparency. Citizens must know how the system works, what counts as a valid evidence, and how errors can be corrected. In Pakistan, this information is scattered, inconsistent or simply unavailable.

Second is accountability. If an AI system incorrectly penalises a citizen, who is responsible? The police? The software vendor? Or the system operator? Regrettably, Pakistan has no law or guideline that defines this.

Third is oversight. AI systems require systematic audits, accuracy testing, fairness assessments and independent third-party evaluations. None of this is reported in the current deployment. Thus, while the e-Challan system may reduce bribery and improve enforcement, it has introduced an algorithmic error, without a vigorous mechanism to detect, explain or correct it.

Although I believe that we must have an aerial view. If a country cannot govern AI in a moderately simple domain like traffic monitoring, how will it manage AI-powered tax audits, judicial queues, border monitoring, welfare targeting or public service delivery?

To address this challenge, policymakers must act decisively on four fronts.

1) Establish a National AI Audit & Accountability Framework to outline acceptable accuracy levels, error thresholds, reporting obligations, mandatory audits and redress mechanisms.

2) Pass a Data Protection and AI Rights Law to give citizens the right to appeal automated decisions, demand evidence, request corrections and challenge unfair or opaque AI outcomes.

3) Mandate public dashboards for AI enforcement, publishing accuracy rates, false positives, disputed challans and reversed decisions. This bold step would build transparency and built public trust.

4) Integrate the system into a National Smart Mobility Plan. Through proper governance, AI enforcement can contribute to accident reduction, congestion management, faster emergency response and stolen vehicle tracking.

AI can issue tickets albeit only responsible governance can build trust. E-Challan rollout is our first real test of AI governance. Whether we pass it depends entirely on how seriously we respond to this moment.