Managing Agentic Security at Scale
Salesforce, Wednesday, October 14th, 2026
Salesforce security explains how it triages the flood of vulnerability findings that frontier models and agentic harnesses now surface.
Salesforce's security team says frontier AI models combined with an agentic harness now surface more real, exploitable findings than engineering teams can absorb, and because discovery has accelerated faster than the rest of the lifecycle, the hard question has moved downstream to prioritization.
Validation still depends on environment access, credentials and expert judgment, and remediation follows established engineering processes.
The post describes how Salesforce decides which findings deserve engineering attention and on what timeline.