Looper

Phone scams cost Austrians alone €20 million in a single year, with 8,600 reported cases — and most fraud never gets reported. The people most affected are the ones least able to protect themselves: elderly parents and grandparents who answer unknown numbers because they're afraid of missing their doctor, their family, or their pharmacy. Today's anti-scam tools all do the same thing: block the call. Scammers don't care. They just dial the next number. Their entire business model depends on volume — thousands of attempts a day looking for one person who picks up.

Looper flips the model. Instead of blocking scammers, we waste them. Looper is an AI agent that picks up scam calls on the user's behalf. It plays a convincing human persona — Gertrude, 74, slightly hard of hearing, owns a cat named Karl — and engages the scammer in a realistic conversation. It loses its glasses. It asks them to repeat themselves. It tells stories about its late husband. It never reveals real information. Every minute the scammer spends with Looper is a minute they're not stealing from someone real. We don't stop scamming by blocking it. We make it economically unviable.

The buyer and the user are different people — and that's by design. The user is someone like Věra, 72, a retired teacher who lives alone and answers every unknown call. She can't be retrained. Blocking doesn't work for her because she still needs to answer the doctor. The buyer is Martin, her 34-year-old son, who installs Looper after she's already lost money to a fake "Microsoft support" call. He sets it up once. She never has to do anything different. This matters because phone fraud isn't just a financial crime — it's a confidence crime. It teaches the most vulnerable people in our society that their own phone is dangerous. We're rebuilding that trust by making the scammer the one who hangs up first.

In 72 hours we built a working prototype combining:

Voice AI with a realistic elderly persona Real-time speech-to-text transcription LLM-based reasoning that detects scam type and adapts the conversation Telephony integration so the AI can answer real inbound calls A scam scenario database with 10 real-world scripts (bank impersonation, fake police, crypto fraud, package delivery, tech support, tax authority, grandchild scam, WhatsApp scam, romance, energy provider) sourced from Austrian and EU consumer protection agencies A "trap question" feature — users set a question only a real caller would recognise as nonsense (e.g. "what's the name of my dog?" when they don't have one). Scammers answer confidently and expose themselves.

We tested the prototype against real scam call transcripts and validated the conversational flow with sample personas.

Three priorities:

Validate with real users. Talk to people over 60 and the family members protecting them. Ship a pilot with 20 elderly users in Vienna. Carrier partnerships. A1 Telekom, Magenta, and Drei are the natural distribution path. Carrier-level integration removes the install friction for the people who need it most. Open the scam database. Community-sourced scam scripts keep Looper ahead of evolving tactics. We'll publish the database under a permissive license so consumer protection agencies and other builders can contribute.