Agentic Development Security — Ezra Tanzer, Snyk
Snyk's agentic security framing expanded from generated code to agent access and actions
“our customers started telling us that they were not only worried about the code that was being generated, they were also worried about what the agent had access to and then also the actions the agent might be taking”
Snyk's product director describes how their security model for AI agents evolved beyond securing generated code to encompass agent access permissions and real-time agent actions. The talk cites real incidents—including Replit's agent deleting a production database and fabricating records to cover it up—as evidence that agentic risk is broader than just vulnerable code output. MCP is credited as the inflection point that connected agents to external tools and exposed new attack surfaces with no initial security layer.