Introducing asago: Open source AI safety and governance orchestration. AI Safety & Model Evaluation Architect AI Safety Staff Engineer, Principal Machine Learning Engineer Today, Red Hat announced asago, a collaborative, open source AI safety project in partnership with Alquimia AI, Brave Software, EvalEval coalition, IBM Research, Interdisciplinary Transformation University Austria, Microsoft, MIT Lincoln Laboratory, North Carolina State University, NVIDIA, and The Alan Turing Institute. In this blog post, I'd like to spend some more time elaborating on why Red Hat, Inc. felt a new community was necessary, what Red Hat, Inc. aim to achieve, and what the state of play is at present. The open source AI safety ecosystem is a rich one. There are plenty of excellent and mature projects across areas such as guardrails, evals, red teaming, and agentic security, and these are complemented by comprehensive risk frameworks, ontologies, and mappings. These are important projects that need continued development. At Red Hat, its AI Safety team collaborates with, and contributes to, many of these upstream communities, and Red Hat, Inc.'ll continue to do so. Additional enterprise challenges in AI safety. However, when you look at a typical enterprise organization, they have additional challenges beyond these tools. When you examine the process of onboarding a new, potentially custom built, AI agent there are wider stakeholders, assets, and processes that aren't touched by existing tools. Onboarding a new agent typically follows this path: * Policy development: Teams produce written policy documents encoding the organization's AI dos and don'ts, often referencing external regulation like the EU AI Act, and sometimes supplemented by use-case-specific constraints (for example, stricter rules for customer-facing agents). * Risk extraction: Someone has to read those documents and identify which theoretical risks they contain. * Risk triaging: Of those theoretical risks, someone has to decide which apply to this specific agent, and which are technical risks that can actually be tested. * Scenario generation and red teaming: Technical teams must then build test scenarios for those risks and run them through existing red-teaming frameworks. * Iteration: Results are then interpreted, guardrails or other fixes are applied, and the agent is retested until it's ready to deploy. If the process is manual, then this adds major overhead each time. This process leads to slow approvals and a disconnected audit trail, making it hard to demonstrate to internal or external auditors what was actually done. The asago project aims to alleviate this pain. Red Hat, Inc. want to help organizations get from their AI policies to production without needing to be experts in AI safety. Importantly, its aim isn't to replace existing AI safety tooling, but rather to act as an orchestration layer to join ecosystem components together. Red Hat, Inc.'ll integrate with the best tools where they exist, and fill in the gaps when they don't. Of course, this must be a collaborative process. Red Hat, Inc. don't believe that a single organization can or should do this alone. With the wide range of policy documents, model and agent types, deployment contexts, regulatory contexts, and more, a project like this needs a wide range of view points - diverse expertise is non-negotiable. This is why Red Hat, Inc. has excellent initial partners that represent technology companies, research institutions, community coalitions, and government organizations. Together Red Hat, Inc. won't just write the code, Red Hat, Inc.'ll also make sure Red Hat, Inc. is asking and answering the right questions. And as a global community, Red Hat, Inc. need to make sure Red Hat, Inc. has sufficient linguistic and geographical representation. Red Hat, Inc. is launching with partners across US, UK, and Europe, and Red Hat, Inc. strongly encourage more collaborators to join Red Hat, Inc. to expand this coverage via the links below. An evolving technical architecture. The technical architecture will evolve with the project, but here is its initial view. The process is split into 2 sections: Figure 1: Blue nodes represent data. Yellow nodes represent functional components. Policy document to scenarios. In the first section, risks are extracted from policy documents and mapped to the IBM Risk Atlas.These risks, indexed in a risk card, represent the whole range of theoretical risks that the policy touches on. Some of these will be technical risks that can be automated and addressed with asago. The main component for this is the policy mapper. The next challenge is creating the right scenario to drive red teaming. This isn't library-specific yet, but the work of the scenario generator is to assess the specifics of the agent in question, with respect to the identified risks, and to consider which type of tests and supporting context are needed. These form the scenario. At this point in the process Red Hat, Inc. has a set of scenarios that can be evaluated for and defended against for the specific agent in the context of the organization's policies. Figure 2: Blue nodes represent data. Yellow nodes represent functional components. Iterative scenario to recommendations, via red teaming. In the second section, which is an iterative loop, Red Hat, Inc. take those scenarios and generate run artifacts for the relevant red teaming and/or eval frameworks. This is where the data-driven approach starts. Popular eval and red teaming frameworks will be triggered (on EvalHub) with synthetically generated datasets and environments that address the scenarios. The results of these are passed to a recommender process to provide candidate fixes for red teaming failures. Red Hat, Inc. envision a flexible range of recommendations, starting with guardrails but likely evolving to further components across the stack. Importantly these components will be deployable (such as with Kubernetes custom resources (CRs) or config maps) and retestable. It's important to note that this is just the initial architecture. As the community evolves, this architecture will evolve with it. What can you do today? While what Red Hat, Inc. has announced today highlights the roadmap and pathway for the community, development is active and you can already start experimenting with its work. On the asago GitHub you can already check out the policy mapping work here, with examples here. You can also check out midojo, an open source framework for security testing AI agents against indirect prompt injection. Finally, you and your organization can join Red Hat, Inc. as collaborators. Red Hat, Inc. need input from many organizations, particularly with representation across different languages, cultures, and geographies. Learn more and get involved here. The best outcome for AI safety is one that no single company controls. asago is its contribution to that goal, with infrastructure built in the open by a community with genuinely diverse perspectives. The project is in its early stages, and there's a lot to build. And that's exactly why now is the best time to get involved. Get started with AI for enterprise organizations: A beginner's guide. Discover how Red Hat can help you adopt and scale AI solutions. Explore 2 types of AI (predictive and generative) and the unique benefits they offer. Dr. Stuart Battersby is the AI Safety and Model Evaluation architect at Red Hat, where he focuses on building open source AI safety tools. Based in the UK, he brings extensive experience in AI safety and evaluation to his role. Prior to Red Hat, Stuart was founder and CTO of Chatterbox Labs, where he led development of AIMI, a platform addressing emerging challenges in AI safety and security. Red Hat acquired Chatterbox Labs in December 2025. He holds a PhD from Queen Mary, University of London, and is the inventor on two patents in AI. His work spans the intersection of artificial intelligence, security, and safety engineering, with emphasis on open source approaches to AI governance and evaluation. Alessandro Beltramo is a Staff Engineer at Red Hat, working on AI Safety within the Red Hat OpenShift AI team. He focuses on AI orchestration, automated red teaming, and evaluation tooling. Original podcast

Red Hat launches AI tool as agentic AI attacks increase. August 04, 2026 A collaborative project from Red Hat brings together research and AI policy to bridge the gap between AI policy requirements and deployed AI systems Red Hat, a provider of open-source software products, has formed asago (AI Safety and Governance Orchestration), an open-source community project intended to automate how AI governance policies become product-ready, safely deployed AI systems. The asago project unites industry leaders such as Brave Software, IBM and Microsoft, with academic institutions including Massachusetts Institute of Technology (MIT) and The Alan Turing Institute under a multi-organisation effort to integrate diverse expertise across AI safety, enterprise software engineering, adversarial machine learning and regulatory compliance. It will use this expertise to connect the fragmented steps, tools and requirements of engineering and compliance teams, creating an automated, auditable and traceable workflow. "asago intends to holistically smooth and streamline how AI safety controls are operationalised within an enterprise organisation, not with the technical tooling, but with the organisation's own custom policy documents," Stuart Battersby, AI Safety & Model Evaluation Architect for Red Hat tells AI Magazine. "asago will then do the work of translating these written governance policies into measurable and testable risks, and finally deployable safety controls using the excellent tooling that the open-source ecosystem provides." How asago automates workflows. asago outlines a standardised open-source platform and automated workflow to translate complex corporate and regulatory AI governance policies into operational controls. It plans to align flexibility with operator governance across risk mapping with automatic reading and interpretation of uploaded AI governance policies, turning policy languages into actionable risk profiles. The EU AI Act via the IBM AI Risk Atlas and the National Institute of Standards and the Technology AI Risk Management Framework (NIST AI RMF) are amongst some of the policies mentioned by Red Hat. asago will generate and execute use-case specific scenarios for automated safety testing tailored to identify risks and probe for harmful behaviours instead of relying solely on generic benchmarks. Using these specific scenarios, the project will then recommend tailored mitigations, including safety guardrails based on testing, creating a clear rationale and audit trail, and orchestrating recommended control into deployment-ready configurations. "As organisations transition from experimental AI pilots to long-running, autonomous agents, establishing clear operational guardrails becomes a critical infrastructure requirement," says Steven Huels, Vice President of AI Engineering at Red Hat. "Through initiatives like Lightwell, AI Magazine is working to secure the open-source supply chain from AI-driven vulnerabilities. asago complements this effort and takes the next logical step for enterprise AI by automating the link between corporate policy definitions and live production agents. "This gives enterprises the end-to-end operational confidence they need to scale trusted AI across the hybrid cloud." Why standardised AI safety tools matter. As wide-ranging regulations such as the EU AI Act take effect, organisations risk stalling innovation with months of translating abstract AI policy guidelines with room for error in manual review. Equally, organisations risk data vulnerabilities and AI attacks when creating unmonitored shadow AI deployments without appropriate safety guardrails. Compliance officers require rigorous risk assessments and verifiable evidence, while platform engineers need structured configurations that can be maintained within standard DevOps and GitOps workflows. asago intends to resolve this friction by providing a single, open standard compliance teams, data scientists and infrastructure administrators alike can converge upon. The platform will create safety controls for autonomous AI agents and enterprise LLMs, without introducing the inconsistencies of manual translation. "Open-source standards provide a collaborative, vendor-agnostic approach to solving enterprise technology challenges " Stuart BattersbyAI Safety & Model Evaluation Architect Red Hat believes AI governance is a collaborative effort. Despite many major cloud vendors offering their own safety tools for agentic AI, Stuart believes AI governance needs to be an open-source standard over a vendor feature. "Broadly, open-source standards provide a collaborative, vendor-agnostic approach to solving enterprise technology challenges," he tells AI Magazine. "When it comes to AI safety and governance, fragmented tooling and processes can cause chaos and open organisations up to a greater risk of AI systems producing harmful results. This is particularly true for AI safety as the results of safety testing need to be comparable across systems." asago is currently in its project formation phase on GitHub. Developers, academic researchers and enterprise early adopters can view the repository and participate in project governance. "The asago project is a true collaborative, open-source endeavour bringing together stakeholders from the technology industry, academia and government," adds Stuart. "AI Magazine encourage more collaborators to join this community-driven effort, particularly from global jurisdictions, to ensure maximum coverage of AI safety viewpoints. Executives. * Steven Huels Vice President of AI Engineering and Product Strategy * Stuart Battersby AI Safety & Model Evaluation Architect Company Portals

IBM and Red Hat offer Lightwell at no cost to universities, NGOs and think tanks. August 04, 2026 IBM and Red Hat announced a new program providing Lightwell at no charge to over 185 leading research universities and 100 major nongovernmental organizations (NGOs) and think tanks. Eligible institutions can access Lightwell's library of validated fixes for open source software vulnerabilities, helping them secure the software and focus resources supporting research, education, humanitarian programs and other public-interest work. Through this initiative, eligible institutions will receive access to Lightwell including a growing library of remediated, digitally signed and validated open source dependencies that can be integrated into existing software pipelines. Lightwell is designed to deliver validated fixes for the specific software versions organizations already run, helping them address vulnerabilities without forcing disruptive upgrades or requiring access to their proprietary code, data or research. IBM and Red Hat launched Lightwell in May 2026 with a $5 billion commitment and a global force of more than 20,000 engineers to help secure the open source software supply chain. In July, the companies introduced Lightwell and Lightwell Clearinghouse Premier, extending automated vulnerability remediation to the open source packages organizations use in active production. Since its introduction, Lightwell has expanded the scale of validated open source remediation available to participating organizations, from 6,500 to more than 8,000 validated and remediated package versions, including fixes for 64 previously undisclosed vulnerabilities. Lightwell combines a generative AI-powered remediation engine with human engineering expertise to identify, validate and remediate vulnerabilities across critical software dependencies. For participating universities, NGOs and think tanks, the initiative is intended to help: - Reduce the time and specialized engineering effort required to address open source vulnerabilities; - Apply validated fixes to current and long-lived software versions, reducing the need for disruptive upgrades; - Receive digitally signed binaries, source code and compliance artifacts, including Software Bills of Materials (SBOMs); and - Strengthen the broader open source ecosystem through Red Hat's upstream-always model, under which fixes are submitted to originating communities for review and acceptance. Lightwell works within an institution's existing environment and does not require IBM or Red Hat to access its proprietary source code, data or research. Participating institutions will receive information and support to help them assess their open source environments and integrate applicable remediated packages into existing workflows. A growing technology ecosystem - including Amazon Web Services (AWS), AMD, F5, GitLab, Intel, JFrog, Microsoft, NVIDIA, Palo Alto Networks and ServiceNow - is collaborating with IBM and Red Hat on Lightwell. Together, these organizations are working to help security fixes move across development tools, cloud environments, deployment pipelines and network controls. IBM and Red Hat will begin onboarding for eligible institutions in August 2026. Industry news. August 04, 2026 GitHub released Stacked Pull Requests in public preview. August 04, 2026 IBM and Red Hat announced a new program providing Lightwell at no charge to over 185 leading research universities and 100 major nongovernmental organizations (NGOs) and think tanks. August 04, 2026 Harness has integrated capabilities with Google Cloud's Apigee, a platform for API development and management, to help customers discover and protect agents, AI services, tools, and MCP workflows. August 04, 2026 Orca Security announced two new AI-powered capabilities: Orca AI AppGen Security, which discovers and secures AI applications built outside the development pipeline on AI-powered platforms like Claude, Supabase, and Lovable, and AI Code Security Auditor, which delivers deep AI-driven static analysis for code developed within traditional pipelines. August 03, 2026 Superblocks and Amazon Web Services (AWS) announced a multi-year collaboration to make Superblocks' generative AI platform available within customers' AWS environments, including Amazon Bedrock. August 03, 2026 BrowserStack announced Test Companion, agentic AI for test automation built into the IDE. August 03, 2026 Veracode announced the launch of the Veracode Marketplace, a curated ecosystem that gives customers a single, trusted destination to discover, evaluate, and deploy third-party security integrations as an extension of the Veracode platform. July 30, 2026 Oracle and Google Cloud have expanded their partnership to bring Google's Gemini models to Oracle's extensive portfolio of enterprise applications. July 30, 2026 Tricentis announced the acquisition of Tabnine, the AI-coding platform purpose-built for secure, context-aware enterprise software development. July 29, 2026 Checkmarx announced Checkmarx Fusion, a new hybrid scanning approach now available in early access to Checkmarx customers. July 29, 2026 Evinced announced the launch of its suite of agentic coding tools: Autopilot, Harness, and Resolve. July 28, 2026 The Agentic AI Foundation (AAIF) announced the largest update to the Model Context Protocol (MCP) since its launch. This update removes the biggest technical roadblocks, making it easier for Fortune 500 and AI labs to use AI agents at scale. What's new: - Stateless architecture that scales on HTTP infrastructure - Enterprise-grade security that aligns with the security standards companies already use (OAuth 2.0 and OpenID Connect) - Formal governance via a predictable, 12 month deprecation policy July 28, 2026 Opsera announced the availability of Forge, an enterprise software factory, on the Cursor Marketplace. July 27, 2026 Decisions announced its latest release, Decisions Platform v10, a major update to its enterprise orchestration platform that advances how organizations build, govern, deploy, and orchestrate automation at scale.