To keep AI under control, companies such as Red Hat are trying to put safeguards in place. The company is now expanding its ability to do so through the acquisition of Chatterbox Labs. Cloud environments, hybrid or otherwise, receive built-in security through Chatterbox’s tooling.
Founded in 2011, Chatterbox Labs focuses on AI security, transparency about AI activity, and quantitative risk analysis. The company’s technology provides automated security and safety tests that generate risk metrics for enterprise implementations. This is an important piece of the puzzle in providing the necessary stability for the advance of AI. IDC predicts AI spending of $227 billion in the enterprise market by 2025, but scaling up pilots to production remains costly and complex. Chatterbox’s protective layer can be particularly useful in taking that step safely.
Three pillars of AI security
The Chatterbox Labs stack rests on three components. AIMI for gen AI provides quantitative risk metrics for LLMs. AIMI for predictive AI validates AI architectures for robustness, fairness, and explainability. Finally, Guardrails identify and remediate unsafe, toxic, or biased prompts before models go into production.
That last feature has become crucial. Many organizations struggle with the question of how to deploy generative AI safely without constant manual supervision. Guardrails technology automates that process and prevents problematic content from reaching the model in the first place. Given the volatile nature of underlying LLMs and the relatively hard limit that the technology seems to reach when it comes to hallucinations, mitigating those problems through stable methods is sorely needed.
Strategic focus on agentic AI
Red Hat sees the acquisition as a logical step toward enriching agentic AI. For example, the Model Context Protocol (MCP) is only intended to allow models to communicate with applications, but does not offer built-in security. Chatterbox Labs has already conducted research into agentic security, such as monitoring agent responses and detecting MCP server action triggers. That technology fits with Red Hat’s roadmap for Llama Stack and MCP support.
Hybrid cloud as a foundation
Red Hat previously announced collaborations with Nvidia, Intel, and AMD to support various AI infrastructures. That strategy of choice over lock-in is now extending to the security side with Chatterbox Labs.