Darktrace Launches AI Security Tool to Tackle Shadow AI Risks

DARKTRACE has launched Darktrace / SECURE AI, a product designed to monitor and govern generative AI use across employees, AI agents and development environments. The company says more than 80% of its customers used generative AI services in August 2026, based on telemetry from about 8,200 monitored deployments, with organisations using an average of five AI providers.

The product applies Darktrace’s behavioural analysis to identify risks that may be missed by static rules, including shadow AI, sensitive prompts, attempted jailbreaks, indirect prompt injection, excessive agent permissions and behavioural drift.

SECURE AI can discover unsanctioned AI services and Model Context Protocol connections, analyse prompts, sessions and responses on supported platforms including Microsoft Copilot, Copilot Studio, ChatGPT Enterprise, Claude, Salesforce and AWS Bedrock, and provide audit trails for agent identities and actions.

Darktrace also says it can help assess agent configurations in low-code and high-code development environments, identify misconfigurations before deployment, and block unauthorised use or quarantine devices through integrations. The product is described as generally available, but the article presents these capabilities as Darktrace’s product claims rather than evidence of a confirmed attack or independent evaluation.