Share with your CISO
Qualys is betting that the real gap in enterprise AI security isn’t model governance on paper, it’s the absence of any visibility into where AI actually runs and what it does at runtime. The company’s TotalAI platform expansion adds kernel-level instrumentation via eBPF, adversarial red-teaming against both large language models and MCP servers, and audit-ready evidence trails mapped to the EU AI Act and OWASP LLM Top 10. The product is generally available now.
What this means for your business
The organizations most exposed here are the ones that have approved AI tools through procurement but have no inventory of what’s actually running, browser extensions, shadow deployments, containerized models spun up by engineering teams without a ticket. If your AI governance program lives in a policy document rather than a detection system, you’re governing a fiction. CISOs who’ve already built out cloud security posture management programs will recognize the pattern immediately, because shadow AI today looks exactly like shadow IT did in 2015, right before the audit findings started arriving.
The technically interesting move in TotalAI is the MCP server scanning, not just access control over MCP. MCP, the Model Context Protocol that lets AI agents call external tools and services, has become the new attack surface that most security vendors are treating as a permissions problem rather than a code vulnerability. Qualys is scanning the server itself for issues like tool poisoning (where a malicious tool description manipulates an agent’s behavior) and SSRF, server-side request forgery, where the agent is tricked into making unauthorized network requests on behalf of an attacker. That’s a meaningful distinction from what the governance-first vendors are shipping.
The falsification condition for TotalAI’s value proposition is whether eBPF instrumentation can actually keep pace with the rate at which new AI deployment patterns emerge, every new agent framework, every new model-serving abstraction adds a layer the kernel probe has to see through. If Qualys can demonstrate coverage fidelity in a large, heterogeneous environment rather than a controlled demo, the CISO who’s been waiting for AI security to mature past checkbox compliance has a credible reason to move a budget line. If it can’t, this is a sharp story about a real problem with a product that still has a gap between its marketing and its detection depth.
Based on reporting from Qualys Announces New Capabilities in TotalAI for Enterprise Ai Security and Governance, originally published 2026-07-29 11:36:00.
