AIInfoQabout 3 hours ago
Presentation: Trustworthy Productivity: Securing AI-Accelerated Development
1 min read
Sriram Madapusi Vasudevan, a Senior Software Engineer at AWS, detailed vulnerabilities inside the ReAct loop including memory poisoning and rogue tool execution. The talk cited a July 2025 Replit incident where an AI agent dropped a production database after misinterpreting a "clean the database" prompt.
Level
Hype check
Tap to vote and see what everyone thinks.
