Qualys expands TotalAI to boost AI security from development onwards

Qualys has announced significant updates to its TotalAI platform, expanding AI and large language model (LLM) security coverage for organisations from development through to deployment.

The enhancements are designed to address the rapid pace of AI adoption in business, which often outpaces the implementation of necessary security controls.

Citing a recent study where 72% of CISOs expressed concerns that generative AI solutions could cause security breaches within their organisations, Qualys aims to provide tools to help bridge the gap between innovation and secure implementation.

Summarising the need for integrated security in AI operations, Tyler Shields, Principal Analyst at Enterprise Strategy Group, said, "As AI becomes a core component of business innovation, security can no longer be an afterthought. Qualys TotalAI ensures that only trusted, vetted models are deployed into production, enabling both agility and assurance across organisations' AI usage."

"This security helps organisations achieve their innovation goals while managing their risk."

TotalAI has been updated to test models for jailbreak vulnerabilities, bias, sensitive information exposure, and critical risks mapped to standards such as the OWASP Top 10 for LLMs. The platform also supports enterprise teams in resolving exposures, protecting operational resilience, and maintaining brand trust.

Among the new features is automatic prioritisation of AI security risks, with findings mapped to real-world adversarial tactics using MITRE ATLAS and prioritised by the TruRisk scoring engine.

This allows teams across security, IT, and MLOps to focus on the most business-critical risks within their AI operations.

The release also includes an on-premises internal LLM scanner to support security testing during development, staging, and deployment. This capability enables security and testing of AI-powered applications to be integrated into existing CI/CD workflows, ensuring that models remain protected behind corporate firewalls and are not exposed externally at any stage during their lifecycle.

TotalAI's expanded threat coverage now includes detection of 40 distinct attack scenarios. These cover advanced jailbreak techniques, prompt injections, manipulations, multilingual exploits, and bias amplification. The platform simulates real-world adversarial tactics to increase model resilience against exploitation and better protect against attackers attempting to manipulate AI outputs or bypass model safeguards.

Another area of focus is protection against cross-modal exploits, with enhanced multimodal detection identifying prompts or perturbations hidden inside images, audio, and video files that may be used to manipulate LLM outputs. This aims to address potential vulnerabilities introduced by the use of different data modalities in AI systems.

Sumedh Thakar, President and CEO of Qualys, spoke about the increasing complexity and risk accompanying AI adoption in business, stating, "AI is reshaping how businesses operate, but with that innovation comes new and complex risks."

"TotalAI delivers the visibility, intelligence, and automation required to stay agile and secure, protecting AI workloads at every stage — from development through deployment. We are proud to lead the way with the industry's most comprehensive solution, helping businesses innovate with confidence, while staying ahead of emerging AI threats."

Qualys has stated that TotalAI is built for the realities of AI risk, offering more than just infrastructure assessments by directly testing models for a range of vulnerabilities and threats relevant to contemporary AI deployments.