Agentic AI is moving from demos to production faster than any enterprise technology in the past. Autonomous systems are already reasoning over proprietary data, calling internal tools, and taking consequential actions inside core business workflows. That shift makes one question urgent: when agents can act, who can inspect, verify, and defend them?
Cloudera's answer is that the defenses protecting our infrastructure must be built in the open. That is why we are proud to be an inaugural partner—alongside NVIDIA and other leaders across cloud, cybersecurity, enterprise software, open source, and AI research—in the newly formed Open Secure AI Alliance, a movement to develop and share open technologies, techniques, and tools to safeguard software and agents in the age of AI.
Why Openness is a Necessity for AI Security
Three reasons this alliance matches our strategy
1. Open Infrastructure is Foundational, Not a Feature
Cloudera has a deep, foundational history in the open-source domain. We have long understood that open data lakehouse formats are a critical pillar of the economy, functioning essentially as a utility that underpins cloud computing, financial services, telecommunications, and government infrastructure.
In the AI era, open-source software and open-weight models are foundational to leadership and cybersecurity. Industry estimates show that between 76% and 99% of commercial software codebases already rely on open-source software.
Open models convert users into builders, accelerating innovation while ensuring that progress isn't restricted to a small number of centralized organizations. More importantly, an open ecosystem allows for independent scientific research and broad scrutiny—enabling defenders to understand, evaluate, and improve how AI systems behave.
2. Neutrality: Any Model, Any Environment
At Cloudera, we champion absolute neutrality. We recognize that to maintain leadership in the AI revolution, the infrastructure running our economy requires safe and secure access to both closed and proprietary models as well as open ones.
Our enterprise customers need the flexibility to adapt AI to their specific industries, proprietary data, customers, and operating requirements. They must be able to do this while protecting their data and avoiding vendor lock-in.
By remaining strictly neutral, we empower organizations to build hybrid systems that reflect their unique priorities.
This is especially vital for highly regulated sectors—such as finance, the public sector, and critical infrastructure—where enterprises must ensure compliance, data sovereignty, and geopolitical resilience, often opting to avoid reliance on APIs hosted abroad. AI safety and defense require a diverse, open ecosystem rather than a single point of control.
3. The Critical Role of the AI Harness in Security and Governance
While the industry has been hyper-focused on foundational models, security requires us to look at the broader picture. An AI agent is not just a model; it is a complex system built from models, harnesses, and guardrails that dictate what an agent can access.
The harness is the critical scaffolding that enables an AI agent to observe context, reason through problems, and take decisive action. Because agents operate autonomously across enterprise systems, securing this harness is paramount. Closed systems risk locking critical intelligence away from defenders. By contrast, open harnesses and tools drastically broaden defensive capabilities and increase transparency.
Through the Open Secure AI Alliance, NVIDIA is contributing vital research to accelerate these cybersecurity tools. For example, the open-source NVIDIA Labs Object-Oriented Agent (NOOA) framework makes it easier to integrate harnesses with models, thereby making agent behavior infinitely easier to test, trace, audit, and govern. These capabilities ensure that when agents are deployed into corporate workflows, their actions remain compliant and observable.
Our agentic portfolio is built for exactly that—running and orchestrating models of the customer's choosing inside their own environment, with lineage, policy, and observability applied end to end. Open inference and gateway controls, a governed context and tool layer, and full auditability mean that when an agent acts, its actions stay compliant, explainable, and inspectable.
Open defense only works if it can be deployed under governance, in production, on data an enterprise cannot afford to move. That deployment surface is what Cloudera brings to the Alliance.
An AI Safety and Security Call to Policymakers—and to Builders
As regulators grapple with AI safety, it is critical that open models, harnesses, and security tooling are recognized as defensive assets, not liabilities. Blanket restrictions on open frontier systems would weaken defensive capacity and concentrate dependence in a few closed providers—the opposite of resilience.
The safer path is the one that gives more defenders the ability to test, verify, and strengthen the systems society relies on, paired with strong safeguards, clear rules against misuse, and rapid remediation.
Openness alone does not guarantee trust. It still takes rigorous testing, secure infrastructure, clear governance, and human oversight. But an open ecosystem is what broadens the community of defenders who can do that work.
Cloudera is committed to working with NVIDIA and our fellow partners to build that future—one where AI is powerful, open, governed, and secure by design. We invite enterprises, governments, and researchers to join us.
