Turkish police AI flags 306,000 Telegram messages; 121 jailed pending trial
Istanbul police built a tool that reads coded language in Telegram groups; 121 suspects are in custody pending trial after a 3 November raid, GamblingNews reported.
Turkish police arrested 169 people in raids across 39 provinces on 3 November, GamblingNews reported, saying police used an AI system built by Istanbul police to find suspects, read coded language and map links between members of alleged illegal betting networks. The reports came out on 7 November, according to the outlet.
Police first targeted 194 suspects. Courts ordered 121 held pending trial, put 45 under judicial supervision and released three without restrictions, according to the report, which cites another outlet’s reporting.
How the tool works
The system is called AVCI, Turkish for “hunter”. It reportedly uses language processing to spot coded language, slang and emoji combinations, and can link accounts under different names. It can allegedly save messages from closed groups as they arrive, even if users later delete them.
In the latest investigation it identified about 306,000 potentially incriminating messages in 47 Telegram groups, the report said, citing another outlet’s reporting. One suspect is a 21-year-old software developer, thought to be a central figure.
Other cases
- A May operation arrested 108 people in 35 provinces after AVCI reviewed around 600,000 messages. It led to 5,151 websites being blocked, the report said.
- A gendarmerie-led probe in Aydın identified 40 suspects whose bank accounts recorded TRY 3.484 billion in transactions. Police said 37 people were detained on 3 November and measures were placed on 467 bank accounts.
- Four suspects were jailed pending trial in October over an alleged scheme using delayed live sports data.
The report does not say whether any of the networks served Turkish users from outside Turkey, or where any operators involved are licensed.
Source
Report by GamblingNewsDrafted with AI assistance from a report by GamblingNews and reviewed by our editors before publication. How we work