Red Hat Brings AI Safety Into the Open With Chatterbox Labs

Red Hat just made a move to strengthen its AI strategy by bringing Chatterbox Labs on board. Chatterbox is known for building safety tools that fit with almost any model, whether it’s generative AI or a more familiar predictive approach. 

The goal is to help companies shift from testing things out to rolling out AI they can trust in real-world use.

“Enterprises are moving AI from the lab to production with great speed, which elevates the urgency for trusted, secure and transparent AI deployments,” said Steven Huels, vice president of AI engineering and product strategy at Red Hat.

Addressing real AI risks across GenAI and predictive AI

Chatterbox Labs offers automated AI testing, factual risk metrics, and a framework that works across any model or architecture, which aligns well with Red Hat’s open, hybrid-cloud approach.

The company has been doing this work since 2011, and they’ve spent that time focused on measuring real AI risk, not the hypothetical kind. Their tools look at fairness, robustness, explainability, toxicity, and whether prompts are safe. 

Because everything they build is model-agnostic, companies don’t have to be tied to a single stack just to validate what they’re deploying.

Here are two of their core capabilities:

• AIMI for generative AI, which provides independent quantitative risk metrics for large language models.
• AIMI for predictive AI, which evaluates model performance against pillars like robustness and fairness.

They also have guardrails that spot insecure or biased prompts before they make it anywhere. Getting that clarity early on is huge for teams focused on responsible AI.

“We cannot allow safety to become a proprietary black box,” said Stuart Battersby, co-founder and chief technology officer of Chatterbox Labs. “It is critical that AI guardrails are not merely deployed; they must be rigorously tested and supported by demonstrable metrics.”

Why this matters for agentic AI and hybrid cloud environments

Red Hat has been building up its platform for agentic AI, including support for the Model Context Protocol (MCP) and the Llama Stack. Chatterbox Labs has already conducted investigative work to monitor agent behaviour and detect MCP server action triggers. 

Bringing those capabilities together gives Red Hat a more complete safety layer for the next wave of autonomous workloads.

The acquisition is also a natural extension of Red Hat AI 3 and the company’s focus on open, hybrid-cloud AI. Red Hat wants customers to run any model on any infrastructure while still ensuring safety from the start. 

Huels emphasized this in his statement: “By integrating Chatterbox Labs into the Red Hat AI portfolio, we are strengthening our promise to customers to provide a comprehensive, open source platform that not only enables them to run any model, anywhere, but to do so with the confidence that safety is built in from the start.”

Folding in Chatterbox Labs basically hands Red Hat users real, production-ready AI with some serious safety metrics.

The acquisition also lines up with what Red Hat highlighted in its AI 3 update. That release focused on getting AI into real production, and Chatterbox Labs adds the safety layer that makes that path more reliable. Together, they show Red Hat tightening its focus on trustworthy, scalable AI.