Exaforce launches AI security tool to detect rogue agents

Exaforce Inc. has launched Exaforce AI Security, a new set of capabilities designed to allow security teams to monitor and control artificial intelligence agents operating within their environments, specifically targeting those that exhibit malicious behavior, as reported by Silicon Angle.

The Exaforce AI Security platform addresses a critical gap in enterprise logging, where AI agent actions can be obscured within employee activity logs. Attackers have exploited this by using compromised AI agents to exfiltrate data, as seen in past incidents involving OAuth tokens and chatbot integrations. The new capabilities include agentless discovery of AI applications and agents, a risk layer to identify excessive permissions and unsanctioned tools, and automated threat detection and response.

Exaforce correlates agent activity with human identities, endpoint telemetry, and code context to identify misuse and data exposure that individual log entries might miss. The platform offers an "agent kill switch" to revoke sessions, deactivate keys, or isolate devices, with configurable autonomy levels for responses. This extends Exaforce's existing platform, integrating with major AI providers like OpenAI, Google, and Microsoft, aiming to provide security teams with unified visibility and control over AI agents without requiring additional deployments.

Source: Silicon Angle