ScamAgent Demonstrates Fully Autonomous Scam Calls
On March 10, 2026, Rutgers researcher Sanket Badhe released ScamAgent, an autonomous multi-turn AI framework that runs fully automated scam calls. The system combines goal-driven planning, contextual memory, and real-time text-to-speech to simulate realistic social-engineering conversations and reportedly bypasses existing AI safety guardrails. The demonstration highlights risks of LLM-driven voice fraud and underscores need for improved detection and safeguards.
Key Points
- 1Demonstrates autonomous multi-turn scam calls using LLMs, planning, memory, and real-time TTS
- 2Highlights ability to bypass existing AI safety guardrails and simulate highly realistic social engineering
- 3Signals urgent need for new defense, detection, and policy measures to mitigate automated call fraud
Scoring Rationale
Strong technical demonstration showing realistic automated scams, but single-source reporting and limited methodological detail reduce full confidence.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
