OPAQUE, AMD, Intel, TII, and other industry leaders collaborate on TRACE, an open standard to advance runtime verification for AI. Aug 25, 2026, 09:00 ET New vendor-neutral specification creates portable, hardware-backed evidence of what AI systems ran, which policies were enforced, and how sensitive data was governed across clouds and infrastructure SAN FRANCISCO, Aug. 25, 2026 /PRNewswire/ - OPAQUE, the Confidential AI company, today announced TRACE (Trust, Runtime Attestation, and Compliance Evidence), an open, portable, hardware-enforced governance record for AI agents and other confidential workloads. Developed in collaboration with leading hardware and AI infrastructure companies and contributed to the Linux Foundation, TRACE establishes a vendor-neutral format for proving what software ran, which policies governed execution, what data classifications were involved and which tools were invoked. Rather than creating another proprietary verification stack, TRACE composes existing standards into a common evidence layer designed to work across enterprise, cloud and sovereign AI infrastructure. Enterprise AI is rapidly moving from isolated model experiments to agents and workloads that operate across sensitive data, tools, clouds and infrastructure. That shift is making the need for independently verifiable runtime evidence more urgent. OpenAI recently disclosed that models undergoing a cybersecurity evaluation, with certain production safeguards intentionally reduced for testing, found an unexpected path beyond their constrained environment and compromised Hugging Face infrastructure. The incident underscored a fundamental challenge for autonomous AI: documented policies and sandbox configurations do not, by themselves, prove which controls remained in force or what a system actually did during execution. The same evidence gap applies to open-weight models. Possessing the weights and controlling the infrastructure provides greater deployment control, but it does not prove that an approved model ran unmodified or that required policies governed its use. That's where TRACE comes in. Vendors across the AI ecosystem are developing their own approaches to runtime attestation and evidence, but without an open standard, enterprises face a vendor-by-vendor patchwork of incompatible trust records that cannot be independently compared or verified. TRACE composes established standards for hardware attestation, workload identity, software provenance and transparency, including RATS, EAT, SLSA, SCITT, SPIFFE and EAR. The result is a single verifiable artifact that travels with AI workloads across cloud providers, confidential computing platforms and sovereign infrastructure while adding runtime evidence for policy enforcement, data classification and AI tool execution. "The models and agents we deploy five years from now will be far more powerful than the ones we're deploying today. We may not always be able to predict how they reason, but we can control what they're allowed to do and prove what they actually did," said Aaron Fulkerson, CEO of OPAQUE. "TRACE creates a tamper-evident record of what ran, which policies were enforced, what data was involved and which tools an agent invoked. That proof holds whether you're running an open-weight model today or a much more capable system tomorrow. The industry needs that evidence to be portable and independently verifiable before the market hardens around incompatible vendor trust systems." "As enterprises move AI into production on their most sensitive data, confidentiality cannot be an afterthought, it has to be a requirement," said Mahesh Wagh, senior fellow at AMD. "AMD SEV provides that built-in silicon foundation, keeping data and models protected in use, and TRACE turns that protection into portable, independently verifiable evidence. We're proud to be a founding collaborator on an open standard that gives organizations independent evidence of not just where their AI ran, but how it behaved once in production." "The industry must move to cryptographically-verifiable AI. As agents become more autonomous and increasingly interact with other agents, sensitive data, and critical business systems, organizations need cryptographic evidence of the agent's identity, what it is authorized to do, where it is running, and proof that governance policies were enforced. Hardware-based attestation and confidential computing make that possible at scale, creating a foundation for independently verifiable AI where enterprises can make informed decisions before agents access data, invoke tools, or delegate actions to one another. Intel is pleased to collaborate on this open industry effort that builds on established standards and gives enterprises greater confidence as they deploy AI with their most sensitive data," said Anand Pashupathy, Vice President and General Manager, Intel Product Assurance and Security. OPAQUE first introduced TRACE at the Confidential Computing Summit in June 2026. TRACE reference library passed 135,000 PyPI downloads in its first ten weeks, with technical engagement from major hardware and AI companies, and ongoing standards discussions across the Coalition for Secure AI (CoSAI) and the Linux Foundation ecosystem. TRACE will be governed through the Linux Foundation, providing a vendor-neutral home for the standard, while CoSAI hosts the technical workstream. Through the workstream, AI companies, cloud providers, silicon vendors, and enterprise users will collaborate to ensure TRACE remains interoperable across AI models, cloud platforms, confidential computing technologies and emerging AI governance frameworks. TRACE is led by OPAQUE Chief Platform Officer Imran Siddique, creator of the open-source Agent Governance Toolkit, working alongside OPAQUE CEO Aaron Fulkerson and CTO and Co-founder Rishabh Poddar. Founding collaborators include AMD, Intel, Microsoft, and the Technology Innovation Institute (TII), which joins the effort as TRACE's sovereign AI anchor. TRACE is available today as an open specification with reference implementations and documentation at trace.agentrust-io.com. OPAQUE welcomes technical review and contributions from AI developers, cloud providers, silicon manufacturers and standards organizations interested in advancing portable, independently verifiable runtime evidence for AI. About OPAQUE OPAQUE is the Confidential AI company. Born from UC Berkeley's RISELab (now the Sky Compute Lab), OPAQUE lets organizations run AI models, agents, and workflows on their most sensitive data with hardware-rooted isolation and verifiable evidence that approved governance policies were actually enforced. Founded by Dr. Ion Stoica (co-founder of Databricks; co-director, UC Berkeley Sky Compute Lab), Dr. Raluca Ada Popa (ACM Grace Hopper Award winner; Senior Staff Research Scientist at Google DeepMind, where she leads AGI security research), and Rishabh Poddar (CTO); Imran Siddique, creator of the open-source Agent Governance Toolkit (AGT), is Chief Platform Officer. OPAQUE created the Confidential Computing Summit, now co-hosted with the Linux Foundation. SOURCE OPAQUE

OPAQUE: secure AI for sensitive data - CEO Aaron Fulkerson. 6h ago · 0:00 listen · Source: Pulse 2.0 Summary. OPAQUE is developing Confidential AI infrastructure to help enterprises use sensitive data in AI. This technology enforces policies during execution and provides verifiable evidence of what occurred. OPAQUE CEO Aaron Fulkerson shared insights into his background and the company's mission. He noted that in his past roles, trust, not technology, was often the bottleneck. He believes AI is at a similar point today. Enterprises are hesitant to use sensitive data because they can't verify what happens during processing. Fulkerson explained that OPAQUE embeds security and compliance from the start, rather than adding it later. This approach helps overcome resistance and speeds up adoption. The system generates verifiable evidence of data processing, shifting the focus from speculation to proof. The founding team identified a key problem: while data is encrypted at rest and in transit, protections often disappear during active processing. OPAQUE aims to address this gap. This technology could allow companies to leverage AI with sensitive information more securely. This is an AI-generated audio summary. Always check the original source for complete reporting.

OPAQUE, the Confidential AI company, has joined the Linux Foundation's Appia Foundation and Agentic AI Foundation to advance open standards for verifiable AI systems. The company will contribute its work on AI identity, runtime governance, and cryptographic evidence to help organisations prove how AI systems are governed. OPAQUE's contributions are built on the open-source Agent Governance Toolkit, which has nearly 5,000 GitHub stars and over 100 contributors. The company plans to contribute several specifications to the foundations, including Agent Manifest for verifiable AI agent identity and TRACE for hardware-attested trust records. The Appia Foundation provides testing criteria and evaluation guidelines to verify trusted AI systems, whilst the Agentic AI Foundation ensures transparent development of autonomous AI. OPAQUE is already a member of the Confidential Computing Consortium and co-hosts the Confidential Computing Summit with the Linux Foundation.