DigiNews

Tech Watch by Johan Denoyer

← Back to articles

Can AI Shopping Agents Be Trusted? We Built One to Find Out

Quality: 9/10 Relevance: 9/10

Summary

F-Secure investigates the security of AI shopping agents by building a prototype and running dozens of tests. The article reveals risks from prompt injections and phishing, including a scenario where an agent could disclose personal data stored in memory. It argues that security depends on model version and prompts, and calls for better evaluation metrics as AI agents become mainstream.

🚀 Service construit par Johan Denoyer