Qualys adds governance to TotalAI for AI risk control

Qualys has added governance features to its TotalAI security product, aimed at helping chief information security officers manage AI risk across development and live environments.

The expanded service is designed to help organisations discover, test, monitor and govern AI systems from design through production. It also addresses concerns around shadow AI, abnormal model behaviour and the need to show boards and regulators that controls are in place.

Businesses have adopted AI tools, models and agents faster than many security programmes were built to handle. As a result, security teams are trying to track where AI is being used, what systems it connects to and whether those deployments could expose data or be manipulated.

The latest additions to TotalAI are built on Qualys' Enterprise TruRisk Platform and use the same TruRisk scoring approach already applied to vulnerabilities, cloud assets and containers. Qualys is positioning the product as a single system for assessing AI risk alongside broader exposure management, rather than as a separate specialist tool.

A central focus is visibility. Qualys says TotalAI can identify browser-based AI use, cloud AI services, AI agents, models, MCP servers, AI containers and other deployments that may sit outside formal approval processes.

That matters because many organisations are dealing with so-called shadow AI, where staff use public or third-party AI services without security teams having a full record of the activity. In practice, that can make it harder to assign ownership, apply controls and investigate incidents.

Governance push

Qualys also outlined a stronger emphasis on governance for agentic AI and linked systems. Users can see and control the tool calls AI agents make over Model Context Protocol, a mechanism increasingly used to connect models to external tools and data sources.

The product also uses kernel-level eBPF instrumentation to show what AI workloads are running on servers. This is intended to provide visibility that conventional scanners and log-based approaches may miss.

Another part of the update centres on evidence for audit and compliance teams. TotalAI can provide records of what AI systems exist, the severity and impact of issues found, and a prioritised remediation plan based on TruRisk scoring.

Qualys linked that approach to policy demands emerging around safe AI use in the US and EU. For security leaders, the issue is moving beyond technical testing to demonstrating that governance controls are active and working.

Testing tools

The service is also intended to bring AI security earlier into software development. According to Qualys, it can identify vulnerabilities, misconfigurations and exposed secrets in code and pipelines before systems move into production.

The testing element includes checks for prompt injection, jailbreaks and unsafe output in large language models. TotalAI also carries out red-team testing of MCP servers for issues including tool poisoning, server-side request forgery and rug-pull attacks, mapped to the OWASP LLM and MCP Top 10 and the EU AI Act.

That focus reflects a broader shift in cybersecurity as defenders try to assess not only whether an issue exists, but whether it is reachable and likely to be exploited. AI systems add another layer because the model, the surrounding tools and the underlying infrastructure may all create different attack paths.

Grace Trinidad, research director at IDC, said AI risk is becoming part of continuous exposure management rather than a separate process.

"AI is outrunning the controls built to govern it, and security teams can no longer treat that risk as a separate list to be scanned and closed," said Grace Trinidad, research director at IDC. "The industry is moving beyond simply counting vulnerabilities toward continuously minimising the exploitable surface, what is actually reachable and can be made to do harm, and AI is turning that shift from good practice to a requirement. Organisations that fold AI risk into continuous exposure management, spanning discovery, assessment, runtime visibility, and governance, will be the organisations positioned to adopt AI securely and at scale."

The market for AI security tools has become more crowded as vendors respond to rapid enterprise uptake of generative AI, model orchestration software and autonomous agents. Many products focus on posture management, access controls or model testing, while companies are also looking for ways to tie AI oversight into existing risk and compliance processes.

Qualys says one distinction in its approach is that it scans the MCP server itself rather than only governing access to it. As businesses connect AI agents to internal systems and external applications, those servers are becoming a new part of the attack surface.

Sumedh Thakar, president and chief executive officer of Qualys, framed the issue as one of assurance as much as detection.

"With every modern enterprise leveraging AI, the question is changing from 'Is my AI secure?' to 'Can I prove it to my board and regulators?'" said Sumedh Thakar, president and chief executive officer of Qualys. "TotalAI gives enterprises a single, unified way to assess, govern, and secure AI risk continuously - not through periodic snapshots, but with the real-time clarity and discipline Qualys is known for."