Red Hat has announced the purchase of AI security company Chatterbox Labsfocused on establishing limits and barriers to AI systems regardless of their models, as well as generative AI. With this operation, Red Hat wants to increase its security implementation capabilities for Red Hat AI products and services, as well as technology relevant to machine learning operations.
The technology developed by Chatterbox Labs provides model-agnostic security testing and guardrails as a critical security layer for AI. This in a framework in which companies are moving from experimentation to production, at which time they have to implement powerful, but also reliable and safe, models.
The Chatterbox Labs integration creates a unified platform where security is built in, strengthening the company’s ability to enable production AI workloads with any model and on any accelerator.
Founded in 2011, Chatterbox Labs has technology and expertise in AI security and transparency. Specifically, it offers automated and customized AI security and protection testing capabilities, providing the objective risk metrics necessary to approve the implementation of AI in production. The technology offers a model-agnostic approach to validating data and models through AIMI for generative AI, which provides independent quantitative risk metrics for LLMs.
Also from AIMI for predictive AI, which validates any AI architecture on key pillars, such as robustness, fairness and explainability. Regarding the deployment of limits and barriers, the company’s technology identifies and corrects unsafe, toxic or biased indications. It also does so before putting the models into production.
This operation is therefore in line with Red Hat’s vision of supporting various deployment models and objectives in the hybrid cloud. Additionally, it complements the future capabilities of Red Hat AI 3, specifically for agentic AI and the Model Context Protocol (MCP).
Chatterbox Labs has also conducted research into holistic agent security, which includes monitoring agent responses and detecting triggers for MSP server actions.
This research is aligned with Red Hat’s roadmap for Llama Stack and MCP support, which will make it easier for the company to secure the next generation of intelligent, automated workloads on a reliable, enterprise-ready foundation. By combining Red Hat’s MLOps capabilities with Chatterbox Labs’ protection capabilities, Red Hat will enable businesses to deploy their AI investments with more confidence.
Steve Huels, Vice President of AI Engineering and Product Strategy at Red Hathas highlighted that «Enterprises are moving AI from the lab to production at high speed, increasing the urgency for reliable, secure, and transparent AI deployments. Chatterbox Labs’ innovative security testing and model-agnostic firewall technology is the critical layer of “AI security” the industry needs. By integrating Chatterbox Labs into Red Hat’s AI portfolio, we reinforce our promise to customers to provide a complete, open source platform that not only allows them to run any model anywhere, but do so with confidence that security is built in from the beginning. This acquisition will help enable truly responsible, industrial-scale AI.«.
As to Stuart Battersby, Co-Founder and Chief Technology Officer at Chatterbox Labshas commented that «As AI systems proliferate across all aspects of business and society, we cannot allow security to become a proprietary black box. It is critical that AI guardrails are not simply implemented, but rigorously tested and supported by demonstrable metrics. Chatterbox Labs has been a pioneer in this discipline from the beginnings of predictive AI to the agentic systems of the future. By joining Red Hat, we can bring these independent and validated security metrics to the open source community. This transparency allows companies to verify security without strings attached, enabling a future where we can all benefit from secure, scalable and open AI.«.
