The Fraud Boxer Podcast No One is Talking about the Downsides of Agentic AI
12 snips
Jan 26, 2026 
Simon Marchand, a 16-year fraud prevention and security veteran now at DataDome, discusses agentic AI and bot threats. He explores how autonomous agents outpace security, create new refund and deepfake risks, and blur signals that defeat old defenses. The conversation covers adopting trusted agents without blocking revenue, edge intent verification, and upskilling teams to handle AI-driven fraud.
AI Snips
Chapters
Transcript
Episode notes
Scale, Not New Schemes, Is The Threat
- Agents massively increase scale and speed of existing fraud schemes, turning manual workflows into impossible volumes for humans.
- The core threat is scale: identical fraud types but thousands of times faster and larger.
Verify Agent Intent At The Edge
- Stop malicious agent activity as early as possible by verifying intent at the edge of your site.
- Track agent behavior throughout the session instead of trusting a single certification or signature.
Don't Rely On Prompts As Proof
- Be cautious about sharing prompts or LLM transcripts for chargeback proofs because prompts don’t prove who initiated the action.
- Rethink protocols that rely solely on prompt submission as definitive evidence of legitimacy.