Securing Software at the Speed of AI
Palantir, Thursday, August 20th, 2026
Palantir's product security team describes building an agentic multi-agent source-code review capability.
Palantir's Product Security Team describes lessons from building an agentic software security strategy.
The team began experimenting with agentic AI across security workflows more than a year ago. By the time Anthropic launched Project Glasswing, Palantir was already running an internal multi-agent review harness with AI security agents specialized in different areas of security expertise.
Anthropic's Mythos model and OpenAI's cyber programme accelerated that work. The team now has operating capabilities for multi-agent source-code review and related security functions.