Beeline Routes 19.1 Million Scam Calls to AI Family Assistants

Russia-based telecom operator Beeline routed 19.1 million suspected scam calls to AI voice assistants this year, with the systems holding conversations for a combined 140,000 hours.

The total is equivalent to about 5,833 days of call time. Beeline’s anti-fraud platform also blocked more than 870 million unwanted calls during the year, but the company did not specify how many of those were scams.

Beeline CEO Sergey Anokhin disclosed the figures Oct. 6 at the Digital Solutions technology forum in Moscow. He described the AI system as a “cyber family” designed to take scam calls away from people and occupy callers’ time instead.

The assistants use voices and personas representing a grandmother, teenager, mother and father. Beeline expanded the system to 10 voices in August, with the assistant selecting a voice based on the number receiving the call and randomly assigning traits such as personality, occupation and hobbies.

Anokhin said the conversations also generate material for retraining the models. The system analyzes a caller’s words, produces a response with a large language model and converts the response into speech. The response process takes less than one second.

Beeline announced in November 2025 that it was testing its cyber grandmother. The company later broadened the service into the cyber family system, which is available to all users through Beeline’s basic free plan.

The same scam call to the same number within five hours is answered by the same AI character. This keeps the character assignment consistent across repeat calls.