Why this transition is a research hazard, not just a memorisation task
The provisions are not a one-to-one renumbering exercise. Some sections were substantively rewritten — the treatment of police custody duration under BNSS being a recent, well-publicised example of a genuine change in the law, not merely a new section number for an old rule. Others were consolidated, split, or given materially different procedural triggers. A precedent interpreting an IPC or CrPC provision does not automatically transfer its ratio to the corresponding BNS or BNSS provision, even when the language looks similar.
This is precisely the kind of error a general-purpose AI model is prone to, because its training data spans years in which only the old codes existed. Ask it a question under the new framework, and it may confidently answer with the old provision, the old section number, or a hybrid of both — fluent, plausible, and wrong.
A practical approach to dual-regime research
Always establish the date of offence first. Before any substantive research begins, fix which regime governs the matter. This single step prevents the majority of transition errors, because it tells you which corpus to search.
Ask AI to confirm the applicable code explicitly. Rather than asking a general question and hoping the tool picks the right framework, specify BNS, BNSS, or BSA by name and ask the tool to confirm the section number against the enacted text — not against a remembered mapping.
Do not assume automatic cross-applicability of old precedent. Where a new provision has been interpreted by a court under the new code, prefer that authority. Where only old-code precedent exists on a point, read the new provision's text carefully before assuming the old ratio survives unchanged — some do, some do not, and the difference matters.
Watch for hybrid citations. A citation that names an old-code section but cites a post-transition case (or vice versa) is a strong signal of confused or fabricated output. Treat it as a flag requiring the same verification any other suspicious citation would demand.
Use grounded tools that are actively maintained. A retrieval-based platform is only as current as its underlying corpus. Confirm that the tool you use is indexing judgments and enacted text under the new codes specifically, not only the legacy provisions it was originally built on.
Why this discipline compounds over time
The transition period will not last forever, but for the next several years, advocates handling matters that straddle the changeover date — or citing older precedent for a proposition that must now be re-anchored in the new code — will face this exact research challenge repeatedly. Building the habit of confirming the governing regime before relying on any AI-surfaced provision protects against an error that is easy to make and embarrassing to explain in open court.
The takeaway
The shift from IPC/CrPC/Evidence Act to BNS/BNSS/BSA is the single biggest live source of AI-research risk in Indian criminal practice today, precisely because it is recent and still unsettled. The fix is procedural discipline: fix the regime first, verify the section against the enacted text, and treat any cross-regime mismatch as a red flag. Platforms built for Indian practice, like LawWorld, keep their corpus current across both frameworks so every citation is anchored to the code that actually governs your matter.
This article is for general information and does not constitute legal advice.
