Cisco has integrated automated red teaming and AI agent monitoring tools into its Cisco AI Defense platform. Algorithmic red teaming explores more than 200 risk categories, including prompt injection and data exfiltration, using multiple AI attackers that adapt their tactics. Monitoring is designed to detect dangerous behavior after deployment, while the platform follows a discover-detect-defend cycle and incorporates threat intelligence from Cisco Talos. Launched in January 2025, the platform received its biggest expansion on February 10, 2026, followed by a free Explorer Edition on March 23 that can evaluate models and agents in 20 minutes. In June 2026, Cisco added agent validation and harness testing targeting agent-specific vulnerabilities, including indirect prompt injection through emails, documents and web pages.
Source: Read the original article
