Figma Deploys AI Agents for Cyber Defense

Figma is using AI agents to strengthen internal cybersecurity, helping its security team investigate alerts, search previous incidents, examine company systems and prepare code fixes while retaining human oversight.

Built around Panther SIEM, the system investigates audit logs across AWS, Okta, GitHub, GCP and osquery, while connecting to more than 100 additional data sources. Agents can also prepare pull requests for security fixes.

Figma says the approach has reduced resolution time for complex alerts by around 70% and cut on-call pages by 20%. Memory plays a central role, allowing agents to learn from previous alerts, behavioral guidance and database structures.

The company says its security agents have also discovered more than 100 previously unknown vulnerabilities, including two critical flaws missed by traditional tools. Its AI code reviewer reportedly achieved 80% precision within a month.

However, Figma maintains strict safeguards, including draft-only agent-generated PRs and controls against sensitive-data exposure. The bigger lesson: AI agents can accelerate cyber defense, but autonomy without human oversight introduces new security risks.