Cisco AI Defense Adds Claude Mythos Red Teaming

Cisco is turning AI security into a continuous operating loop: attack-test agents before deployment, observe them in production, and enforce guardrails when behavior drifts.

• –Claude Mythos brings stronger adversarial reasoning to multi-step attack planning and payload generation.
• –Authorized targets, named operators, fixed objectives, and audit trails reduce the risk of misusing a powerful cyber-capable model.
• –Splunk Agent Observability connects traces, tool calls, errors, business outcomes, and token costs for operational debugging.
• –The combination addresses both sides of agent risk: malicious influence and ordinary workflow failure.
• –Availability remains limited, with Claude Mythos red teaming coming soon to select enterprises.