Danish SMS fraud filter gave employees access to private messages
An AI-powered SMS filter deployed by Danish telecommunications companies to block fraud messages inadvertently allowed employees to read thousands of private text messages for months. The filter was overly broad and caught many legitimate messages before being partially disabled in July 2025.
What happened
Employees at Telenor, 3, TDC and Norlys were able to read thousands of private SMS messages through an AI-powered filter deployed in spring 2025 to block fraud messages. The filter was designed to analyze all incoming SMS messages using algorithms and artificial intelligence to identify fraud attempts, such as messages sent to thousands of recipients containing malicious links.
The filter proved too imprecise and flagged many legitimate messages along with fraudulent ones, inadvertently exposing private communications to telecom employees. Around July 2025, the companies stopped using the most invasive parts of the filter, which subsequently became less effective.
Reported impact
Thousands of Danish consumers had their private SMS messages accessed by telecommunications company employees without authorization. The exposure violated the constitutional protection of mail and message secrecy and potentially conflicted with EU privacy directives.
Index analysis
The SMS filter was unable to distinguish between fraudulent and legitimate messages with sufficient accuracy, causing the system to flag and make visible a large volume of ordinary private communications. According to the Teleindustrien director, all telecom companies had overestimated the system's capability to filter independently.
Governance controls
The filter was presented with backing from the government, telecom industry and consumer group Forbrukerrådet Tænk in November 2024. Officials stated the system would not store or log data and would require a court order to access private communications, aligned with constitutional and EU privacy protections.
Digitaliseringsstyrelsen is conducting an evaluation of the SMS filter, with results to be discussed with parties in the Danish parliament.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.