Romania's DNSC warns of AI-generated fake police in influence campaign
Romania's cyber security authority warned of an influence campaign using AI-generated characters falsely posing as police or gendarme officers to target social media users through micro-targeted posts. The campaign aims to test public reaction and behavioural patterns to refine future influence operations, potentially associated with personal data collection or fraud.
What happened
Romania's National Directorate for Cyber Security (DNSC) warned on 24 March 2026 of an influence campaign using AI-generated characters falsely presented as employees of the Romanian Police or Romanian Gendarmerie, targeting Romania's population through micro-targeting on social media.
The posts, which appear to reflect real situations, aim to test public reaction and identify online behavioural patterns to calibrate future influence messaging. The campaign exploits emotions and social context to determine what type of content resonates quickly and efficiently with Romanian social media users.
Reported impact
DNSC noted that similar tactics are used in other countries to understand what content 'works' with the public in order to build more efficient influence campaigns in the future.
In some cases, these tactics are associated with attempts to collect personal data or online fraud. DNSC advises users not to distribute suspicious content even if it appears credible, to verify information from official sources, and to analyse images carefully as AI-generated content often contains unnatural details.
Index analysis
DNSC identified that the campaign uses AI-generated characters with increased credibility through false attribution to official law enforcement agencies. The posts are designed to test audience response and identify behavioural patterns to enable more targeted future influence operations.
Governance controls
DNSC recommends users verify information from official channels, analyse images carefully for unnatural details, pay attention to inconsistencies in text, voice or character behaviour, and seek confirmation from multiple independent sources.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.