AI as crime fighter

Massive use of technology, and digitization of records in the government sector along with the introduction of AI agents has resulted in a paradigm shift in the nature of governance ~ from a direct service provider to a manager of complex, automated ecosystems. Slowly but gradually, a revolution is taking place. The concept of a Digital Citizen has crept in, and the “identity” of the governed is increasingly tied to digital footprints. The Government is rapidly moving towards biometric-integrated AI systems to manage social welfare schemes as well as national security.

AI agents can now predict citizen needs or risks, based on historical data. While this enables “proactive governance,” it risks reducing a person’s identity to a data point, which is vulnerable for theft and misuse as well by the interested parties. For the government, maintaining control over the identity data of its citizens is the new frontier of national sovereignty, especially when the AI models used are developed by foreign corporations. Authority is gradually being redistributed from bureaucratic discretion to algorithmic logic. This changes how laws are enforced and how policies are designed. AI agents can monitor financial transactions for money laundering or track cross-border movements in real-time.

Advertisement

This provides the state with “omnipresent” authority that is faster than any human committee. However, if a government agency denies a permit, or flags a transaction based on an AI’s suggestion, the authority of that decision rests on the machine’s “black box.” To maintain public trust, the state must ensure that officials have the final authority to override or explain machine outputs. When AI is integrated into governance, the traditional chain of command, where a minister or official is responsible for an error, becomes complicated and sensitive.

If an AI-driven security system at a border makes a categorical error, or a tax-assessment AI incorrectly penalises thousands of citizens, the “Accountability Gap” becomes a crisis situation for the government. To mitigate that, the governance frameworks are now focusing on Explainable AI (XAI). This requires that every government AI agent must be able to “explain” its reasoning in human-understandable terms, ensuring that an official can be held accountable for the final outcome. Applying the concepts of identity, authority, and accountability to India’s national security landscape reveals a profound technological transformation.

The Ministry of Home Affairs has accelerated its pivot toward an AI-augmented state, reshaping how India monitors and ensures security of its borders, and combats illicit financial flows, in and out. The human-AI dynamic in these high-stakes domains operates under specific, modern frameworks. The management of India’s borders, particularly the complex, riverine terrains of the India-Bangladesh border and the sensitive geometry of the India-Pakistan border, is transitioning from manual patrolling to an algorithmic grid. In border management, “identity” becomes a challenge for rapid categorization of individuals. The Smart Border Project rollout integrates AI-based thermal cameras, radar systems, and smart fencing to establish an “impenetrable” perimeter.

AI agents do not just look for physical fences, they also analyse behavioural anomalies. Along unfenced riverine stretches, such as the shifting courses of the Padma and Brahmaputra rivers, India relies on Border Electronically Dominated QRT Interception Technique. Here, laser sensors and AI analytics determine whether a heat signature is a stray animal, a local fisherman, or an attempted infiltration. With a massive surge in drone-based narcotics and weapons smuggling, especially on the western front, AI agents are critical for identifying aerial signatures, instantly distinguishing bird migration patterns from low-flying hostile payloads, which triggers automated anti-drone RF jammers.

The redistribution of authority on the border introduces a core operational doctrine wherein technology is a force multiplier and not a decision-maker. AI agents handle the authority of detection, and process terabytes of real-time sensory data that would blind human operators. However, the authority of kinetic response remains strictly human. When an AI agent flags a subterranean vibration sensor indicating tunneldigging, the system alerts an Integrated Command and Control Centre. The choice to deploy a Quick Reaction Team (QRT), or use lethal force, is explicitly withheld from the machine. In such a scenario, the accountability will always be a concern.

If an automated laser barrier or drone defence system misinterprets a target, the political and geopolitical stakes are massive. If a tactical field commander acts on an AI alert that turned out to be a system glitch or a spoofed signal from an adversary, accountability rests heavily on the human commander. To address this, the government mandates data logging within the Comprehensive Integrated Border Management System (CIBMS). Every algorithmic recommendation is recorded, ensuring a strict forensic trail for the Border Security Force (BSF) if a cross-border incident occurs.

In anti-Money Laundering (AML) efforts as well, the government has gone into massive algorithmic financial warfare. India’s economic sovereignty is heavily guarded by financial intelligence, where human investigators are being paired with AI agents to track increasingly sophisticated financial crime. Identity of a culprit in this area is dependent on entity resolution and digital shadows. Laundering networks actively split, manipulate, and disguise identities using shell companies, mule accounts, and cross-border crypto wallets. The Financial Intelligence Unit (FIU) and the Enforcement Directorate (ED) deploy AI agents capable of Entity Resolution.

AI ingests fragmented data across the centralized NATGRID platform (which integrates banking records, tax files, immigration logs, and property details).It identifies that five seemingly unrelated companies across different jurisdictions share the same underlying digital signature, phone number, or erratic transaction pacing, unmasking the “true identity” of a laundering syndicate. Traditionally, a human investigator required initial leads or an FIR to demand bank records. AI has shifted this authority dynamic. Predictive Investigation has been resorted to.

AI agents within NATGRID process massive datasets to generate real-time risk scores before a formal case is even filed. According to recent ED compliance data, utilizing AI behavioural analytics has slashed complex transaction tracking timelines from 3-4 years down to just 1-1.5 years. The machine holds the authority to “join the dots” proactively, surfacing hidden structures like massive financial scams or terror financing networks that humans cannot spot. The flip side of the use of AI in financial investigations is the possibility of generating “false positives” that can freeze legitimate corporate assets or citizen accounts, creating a severe accountability gap. To obviate this and to ensure structural accountability, the Prevention of Money Laundering Act (PMLA) frameworks ensure that Provisional Attachment Orders (PAOs), or the freezing of assets, cannot be executed by an AI.

It has to be verified independently by an officer, who will verify AI’s data trail, sign the warrant, and present the evidence to an Adjudicating Authority. AI acts as the investigator’s lens, but the legal and moral liability remains entirely on the human institution. Whether securing a physical riverine border or a digital financial corridor, India’s approach to AI governance underscores a vital lesson: The machine provides clarity and speed, but the state’s legitimacy relies on humans retaining the burden of final judgment.

RAKESH ASTHANA

The writer, a retired IPS officer, has served in various capacities including as Commissioner of Delhi Police, DG-BSF, DG-NCB, DG-BCAS and Special Director, CBI