Exclusive: Union HM Amit Shah's Push Towards AI in intelligence is a must, but the IB's Real Test Is Judgment
Union Home Minister Amit Shah's call at the 9th National Security Strategy Conference (NSSC) urging AI-led intelligence analysis and pre-emptive action against multi-domain threats has been welcomed across India's intelligence power corridors. But for the Intelligence Bureau (IB), the endorsement comes with a caveat: technology can accelerate the intelligence cycle, but it cannot yet replace the judgment at its heart.
Analysis, Not Just Collection
The intelligence cycle is often reduced, in public discourse, to "collection." In reality, its most decisive phase is analysis, the point at which raw data is converted into judgment, and where feedback loops correct course for the next cycle.
According to a top source in the intelligence establishment, IB's HUMINT (human intelligence) and analytical functions carry a subtle but persistent flaw: a tendency toward bias and overconfidence, rooted less in dishonesty than in insufficient depth of strategic knowledge and tradecraft. Some bias, the source noted, is an inherent part of intelligence judgment; no analyst is a blank slate. The concern is not bias itself, but unexamined bias compounding under pressure.
IB, top sources say, is "quietly" upgrading its analytical muscle. But there's an internal debate brewing: will AI induction reduce this deficiency, or simply mask it with a new, more dangerous dependency on automated output? The consensus among sources is clear that AI should sharpen human tradecraft, not substitute for it.
The Cautionary Tale: CIA's AI Overreach
Washington offers a live case study in what happens when this balance tips too far. The CIA's AI-driven intelligence architecture has dramatically scaled up its collection and compilation capacity, but at a cost. Top sources describe a pattern of mismatched correlations and analytical overconfidence, where AI systems struggle to parse complex, layered intelligence that demands contextual and historical judgment , not just pattern matching. The problem with ambitious AI-led intelligence and battlefield AI (Project Maven) is a key example , according to top sources.
The Lesson
The lesson, as one source put it bluntly, is that complex intelligence analysis demands a grasp of the strategic environment and the historical forces that shaped it. This is perhaps the most important insight from Shah's push and IB's internal recalibration: AI buys speed, not necessarily accuracy. Human judgment remains, in the words of one source, "the gold standard" of intelligence, not because machines can't process faster, but because interpretation of intent, deception, and context is still fundamentally a human craft.
Top sources said if there's one arena tailor-made for testing AI's analytical promise before it's trusted with sensitive HUMINT judgments, it's OSINT (open-source intelligence). Making it the ideal proving ground for AI models to be stress-tested, refined, and validated before their conclusions inform higher-stakes decisions.
For IB, the mandate from Shah's speech is an opportunity, not a shortcut. The path ahead lies in a calibrated fusion of HUMINT and TECHINT where AI extends human capacity rather than replacing human discernment.
