Red Hat, NVIDIA, and IBM Join Forces to Transform AI Policy into Actionable Code
In a world where the demand for responsible AI integration grows ever stronger, Red Hat has unveiled **asago**, an innovative open-source initiative designed to transform AI governance policies into actionable deployment code. This groundbreaking project addresses a pressing need: balancing AI innovation with stringent compliance requirements. As organizations grapple with regulatory pressures, asago offers a pathway that streamlines the complex journey from policy to practice.
What is asago?
asago positions itself as an automated, auditable workflow that effectively bridges the gap between the disparate steps, tools, and requirements of both engineering and compliance teams. With regulations like the **EU AI Act** now in effect, organizations face a critical decision: either slow down AI progress with cumbersome manual reviews or risk deploying ungoverned agents without proper oversight.
The Foundation of Collaboration
This initiative springs from a broader effort involving collaboration between Red Hat and NVIDIA within the **Open Secure AI Alliance**. Released under the **Apache License 2.0**, asago is currently in its formative stage. Developers, academic researchers, and early adopters from enterprises are invited to review and contribute to the governance of this evolving project via its GitHub repository.
Four Stages from Policy to Operational Control
Red Hat outlines a **four-stage workflow** essential for integrating governance seamlessly into AI deployment:
- Risk Mapping: The framework begins by interpreting an organization’s governance policy to create a customized risk profile. It aligns specific requirements with established standards such as the **NIST AI RMF** and the **OWASP LLM Top 10**.
- Risk Assessment: Instead of standard checklists, asago generates scenarios tailored to unique use cases, delving into potentially harmful behaviors identified in earlier stages.
- Risk Mitigation: Based on the testing outcomes, the system suggests well-defined guardrails and builds a rationale trail ready for scrutiny.
- Deployment Configuration: asago compiles these recommendations into deployment-ready configurations suitable for both hybrid cloud and Kubernetes environments, aiming to reduce deployment timelines from months to just days.
The Importance of Audit Trails
A standout feature of asago is its commitment to producing a continuous audit trail throughout each stage. This ensures traceability, allowing any active control in a live deployment to be linked back to the corresponding policy clause justifying it.
Red Hat emphasizes that AI safety should not be viewed as a one-time certification but as an ongoing operational necessity. Steven Huels, Red Hat’s VP of AI Engineering, captured this notion succinctly: “As organizations transition from experimental AI to long-standing autonomous agents, establishing clear operational guardrails becomes essential.”
A Collaborative Effort
Red Hat’s **asago** initiative boasts a rich tapestry of contributors extending beyond its founding team. Notable organizations and institutions like **Brave Software**, **IBM Research**, **Microsoft**, and **MIT Lincoln Laboratory** are involved, illustrating that the complexity of AI safety challenges requires a united front.
Sarah Bird, Chief Product Officer for Responsible AI at Microsoft, aptly stated, “Many of the hardest AI safety and security challenges are still unsolved, and no single organization can tackle them all alone.” This sentiment resonates with leaders from academia as well, such as NC State’s Veena Misra, who notes that AI safety transcends policy, positioning it as a vital engineering endeavor.
Looking Ahead with asago
The outputs of **asago** aim to be infrastructure-agnostic, providing configurations that work seamlessly with platforms like Kubernetes, Terraform, and Ansible. However, it’s important to note that the project is still in its infancy. Currently, there are no real-world deployment case studies or benchmarks validating the ambitious claims regarding expedited deployment timelines during regulatory audits.
As asago stands ready for community engagement, it invites developers, researchers, and enterprise teams to contribute to its evolution. By fostering collaboration through its **GitHub repository**, asago opens the door to shared insights and progress in AI governance.
We’re excited about the potential that asago holds for transforming AI governance. If you’re passionate about driving change, why not dive into this initiative? Your expertise could play a crucial role in reshaping the future of AI compliance and safety. Join the conversation and be a part of this important journey!
