The Digital Security Act (DSA) was passed by parliament in October 2018 and replaced the older ICT Act's much-criticised Section 57. Critics warned it would broaden police powers around online speech; supporters argued it consolidated and modernised the country's cybercrime framework. The BPO record doesn't settle that argument. What it does do is show the change in the police interaction pattern starting almost immediately after the law took effect.
The arrest-count line
Between January 2014 and the DSA's passage in late September 2018, the BPO record contains 59,693 coded incidents, with 148,261 cumulative arrests across all incident types. In the comparable post-DSA window (October 2018 through December 2024): 81,741 incidents and 193,511 arrests.
Two windows of similar length, similar incident volume, substantially different arrest-count shape. Not all of that delta is DSA-driven — the period also covers two general elections, a global pandemic with associated curfew enforcement, and the ongoing Rohingya operation. But the post-DSA arrest curve is visibly steeper, and its actor composition has shifted.
What kind of arrests
Before the DSA, arrests in the BPO record skewed toward two archetypes: street-level enforcement (drug raids, hartal-day detentions, smuggling busts) and post-election sweeps. After the DSA, a third archetype became prominent: arrests of journalists, activists, students, and ordinary citizens following a complaint filed under Sections 25, 28, 29, or 31 of the new act. These incidents appear in the BPO record under various violence-type codes (Coercion, Intimidation, Politically motivated detention), with eventurl pointing to a press report.
Crucially, the volume of these journalist/activist arrests is small in absolute terms — most arrest-counter events in the BPO record continue to be the older street-level archetypes. But the post-DSA period contains an entire class of incident that essentially did not exist before. That's the structural finding.
What the BPO record cannot tell you
Two things this Story cannot answer from the BPO data alone:
How many cases. The BPO records incidents, not court cases or prosecutions. A single arrest can lead to multiple case filings, dropped cases, or none. For that question, you need separate court-records data (which the BPO desk has begun to collate for the post-2020 period; see the Methodology page for the data-coverage gaps).
Self-censorship. The DSA's most consistent criticism is that it produced a chilling effect on press and online speech. By definition, that effect is invisible in an incident-coded dataset — the speech that doesn't happen leaves no row in the table.