Connect with us

Technology

Sondera Compiles Natural-Language Rules into Provable Control Over AI Agent Actions

Published

on

Sondera’s autoformalization research was accepted at ICML 2026’s Agents in the Wild workshop and at FLoC 2026’s LLM-Solve workshop, with a related tool demo at Black Hat Arsenal. Sondera transforms natural-language policy into formally verified rules that run on any agent.

NEW YORK, June 30, 2026 /PRNewswire/ — Every organization runs on rules, security, compliance, and business logic written in natural language. As AI agents take on real work, those rules have to bind them too. Sondera today announced that its research on compiling natural-language policy into formally verified controls for AI agents has been accepted at workshops at ICML 2026 and at the Federated Logic Conference (FLoC) 2026’s LLM-Solve workshop, and selected for a tool demonstration at Black Hat Arsenal.

The paper, “Autoformalization of Agent Instructions into Policy-as-Code,” by Sondera’s Adam Mondl, Matthew Maisel, and John Brock, was accepted at the Second Workshop on Agents in the Wild at ICML 2026, one of the field’s leading machine learning conferences, and LLMSolve, a workshop at FLoC on applying large language models to formal and solver-based reasoning. A related tool, “GolemHalt: A Deterministic Reference Monitor for AI Coding Agents,” will be demonstrated at Black Hat Arsenal, Black Hat’s curated showcase of open-source security tools.

In peer-reviewed research using MedAgentBench, an independent benchmark for clinical AI agents published in NEJM AI, Sondera’s pipeline autoformalized more of the 88-rule policy than prior published work had hand-coded (23 of 88), and the resulting rules blocked every adversarial unsafe write attempt (99 of 99).

Autoformalization is critical for enforcing agent policies because real policy and business logic live in natural language — in documents like a HIPAA manual, FINRA guidelines, a standard operating procedure, or an agent’s own instructions. Traditionally, turning natural-language policies into something a machine can enforce has meant hand-coding rules one at a time, which does not scale to how fast agents are being deployed.

Sondera’s pipeline reads the natural language and compiles it directly into formally verified Cedar policy-as-code, with a theorem prover checking every rule and adversarial simulation stress-testing it before production, both to find edge cases and to confirm that legitimate work is still permitted. The approach is neurosymbolic: neural classifiers such as LLMs-as-judges evaluate what an agent is doing probabilistically, while symbolic rules decide deterministically what the agent is allowed to do.

At runtime, a verified deterministic rule returns a decision (such as allow, deny, or escalate) for each agent action. Because the policy-as-code is enforced deterministically outside the context window, rather than relying on an AI-as-judge, issues such as prompt injection, emergent behavior, and model drift cannot talk the policy enforcement layer out of its rules. And because enforcement is stateful, tracking the agent’s full trajectory and the flow of information across it, the same action can be allowed or denied depending on what the agent did earlier in the run.

“The agent incidents we see today aren’t from prompt injection and hijacking. They’re from authorized humans asking authorized agents to do legitimate tasks, like analyzing a financial file or configuring a server. Along the way, the agent reaches the goal with unintended behavior, like leaking or destroying data,” said Josh Devon, co-founder and CEO of Sondera. “Even beyond that, enterprises are struggling to apply business logic at scale to their agents, rules that aren’t security or compliance but just standard operating procedure, like a coding agent that should stand up a server on the company’s approved cloud account rather than an unapproved vendor. Our research is focused on letting organizations turn on the most powerful, long-running agents possible while having the confidence that they will follow the rules.”

Sondera’s policy and agent control plane is currently in private beta. Teams interested in participating can visit sondera.ai.

The full paper “Autoformalization of Agent Instructions into Policy-as-Code” is available at arxiv.org/abs/2606.26649.

Sondera’s open-source harness and SDKs are at github.com/sondera-ai.

About Sondera

Sondera gives enterprises provable control over AI agents, in natural language. Its approach is neurosymbolic: neural classifiers detect what an agent is doing, symbolic rules decide what it is allowed to do. Through a process called autoformalization, Sondera compiles an organization’s natural-language rules into formally verified Cedar policy-as-code, checked by a theorem prover and enforced on every action an agent takes, on any agent runtime or harness. Because symbolic enforcement runs outside the model and across the full run, prompt injection and drift can’t bypass it, and the same action can be allowed or denied based on what the agent did before it.

Teams use Sondera to apply complex business logic, security, and compliance rules to coding agents, the agents they build, and the agents they sell. Open-source harness at github.com/sondera-ai. Learn more at sondera.ai.

Contacts
Media Contact
pr@sondera.ai

View original content to download multimedia:https://www.prnewswire.com/news-releases/sondera-compiles-natural-language-rules-into-provable-control-over-ai-agent-actions-302814775.html

SOURCE Sondera, Inc.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Technology

CAI Closes Recapitalization with JLL Partners, Unlocking New Opportunities for Growth

Published

on

By

New partnership positions CAI to accelerate its global expansion, technology investment and enhanced service offerings

INDIANAPOLIS, July 1, 2026 /PRNewswire/ — CAI, a professional services firm dedicated to accelerating operational readiness and excellence in life sciences and mission-critical environments, today announced that it has closed its recapitalization transaction with JLL Partners (“JLL”), a New York-based middle-market private equity firm focused on investing in the healthcare, industrials, and business services sectors.

The closed transaction builds on CAI’s recent momentum, including expanded service for global customers in Australia and new technology partnerships. With JLL’s support, CAI expects to accelerate investments in talent, training, next-generation technologies, strategic acquisitions and enhanced service offerings, while continuing to build on its culture as an employer of choice for technical professionals serving highly regulated and mission-critical industries.

For more than 30 years, CAI has provided operational readiness services to critical industry customers with the technical rigor, precision and consistency their environments require. The partnership with JLL is expected to build on the deep technical expertise and long-standing customer relationships of CAI’s more than 700 professionals operating across North America, Europe, Australia and Asia, while supporting CAI’s continued investment in its people, platform and global capabilities.

“With the transaction closed, we have accomplished what we set out to do: forge a strong partnership with JLL that will enable us to enhance and expand our service offerings as we explore new technologies, geographies, and services,” said CAI CEO Sheena Dempsey. “This partnership is grounded in JLL’s support for our business strategy, our people and our vision for the future.”

Phil Pursifull, CAI CFO, added, “I have been part of CAI for nearly three decades, and this is one of the most exciting moments in our company’s history. We are seeing strong customer demand, continued advancement across technologies and the markets we serve, and tremendous opportunity to invest in the people and culture that have always set us apart.”

NewVale Capital, a growth equity firm focused on investing in life science services companies, was an investment partner in the transaction.

Bourne Partners and Stifel acted as financial advisors to JLL Partners, and Stout acted as financial advisor to CAI.  

About CAI 

CAI is a professional services firm composed of engineering, quality, and operations experts dedicated to accelerating operational readiness and excellence in life sciences and mission-critical environments. With deep roots in CQV (commissioning, qualification, validation), CAI has evolved over the past 30 years alongside the industries it serves, helping organizations bring complex facilities, systems, and technologies online with confidence. 

The elite team at CAI combines deep technical expertise with proven processes and modern tools to deliver projects on time and within budget, supporting outcomes that matter in highly regulated environments. Over three decades, CAI has embraced new modalities and guided clients through digital transformation while strengthening operational readiness. As industry demands grow, CAI strives to define the next era of agility, resilience, and operational excellence. 

For more information, please visit https://caiready.com/

About JLL Partners 

JLL Partners, LLC is a middle-market private equity firm with over three decades of experience transforming businesses in the healthcare, industrials, and business services sectors. The Firm is dedicated to partnering with companies that it believes it can help build into market leaders through a combination of strategic mergers and acquisitions, organic growth initiatives, and operational enhancements. The JLL Partners team is comprised of seasoned investment professionals and operating partners who are focused on driving long-term value creation across its portfolio. Since its founding in 1988, JLL Partners has invested approximately $9 billion of equity capital with 61 platform investments. 

For more information, please visit www.jllpartners.com

About NewVale Capital

Founded in 2021, NewVale Capital is a sector-specialist pharma services growth equity firm investing in proven, revenue-generating services and enabling technology businesses serving the life sciences ecosystem.

For more information, please visit www.newvalecapital.com

View original content to download multimedia:https://www.prnewswire.com/news-releases/cai-closes-recapitalization-with-jll-partners-unlocking-new-opportunities-for-growth-302816296.html

SOURCE CAI

Continue Reading

Technology

SIGGRAPH 2026 Explores the Future of Robotics Through Computer Graphics, Simulation, and Creative Expression

Published

on

By

The world’s premier conference on computer graphics showcases how the technologies behind film, games, and visual effects are shaping the next generation of robots

LOS ANGELES, July 1, 2026 /PRNewswire/ — Computer graphics and robotics have been converging for years. At SIGGRAPH 2026, that convergence takes center stage as robots learn, move, create, and interact through technologies originally developed for film, games, and visual effects. Taking place 19–23 July at the Los Angeles Convention Center, the conference shows how advances in computer graphics are helping shape the future of intelligent machines, revealing a future where virtual worlds are increasingly used to train, test, and inspire robotic systems.

At SIGGRAPH 2026, robotics is not confined to a single program. It runs as a common thread across the conference, appearing in Technical Papers, Emerging Technologies, Courses, Frontiers, Spatial Storytelling, Art Papers, Talks, and Technical Workshops. Together, these sessions demonstrate how the same technologies used to create digital characters, immersive worlds, and visual effects are increasingly being used to design, train, and deploy robots capable of operating in complex real-world environments.

“The lines between computer graphics, physics, and AI are blurring. Because of that, you’re seeing robotics become more pervasive at the conference because these fields are naturally becoming more intertwined,” said SIGGRAPH 2026 Conference Chair Chris Redmann. “It opens up new horizons and pathways for computer graphics research and new modes of interactivity, where the physical world and the digital world become even more complementary.”

Training Robots in Virtual Worlds

A central theme across this year’s robotics programming is simulation-first development, in which robots are designed, trained, and validated in virtual environments before entering the physical world. A trio of NVIDIA-led courses anchors the thread: “How To Build End-To-End Physical AI Systems for Robots” covers data generation, training, and edge deployment for humanoid and general-purpose robots; “Accelerate Robot Learning With NVIDIA Isaac Lab and Newton” introduces a GPU-accelerated physics engine for training and evaluating robot policies; and “Simulating a Dextrous Hand For Robotics With OpenUSD” walks through preparing simulation-ready robot assets.

The same ideas drive a strong Technical Papers showing, where “SimArt: Decomposing Monolithic Meshes into Sim-ready Articulated Assets via MLLM” converts static 3D meshes into simulation-ready articulated assets, “MotionBricks: Scalable Real-Time Motions with Modular Latent Generative Model and Smart Primitives” generates real-time motion for animation and robotics, “ReActor: Reinforcement Learning for Physics-Aware Motion Retargeting” from Disney Research uses reinforcement learning to retarget human motion onto humanoid and quadruped forms, and “Computational Design of Terrestrial Robots with Anisotropic Friction”, from teams including Carnegie Mellon University, Genesis AI, Tsinghua University, and Shanghai Qi Zhi Institute, co-designs robot bodies and controllers for locomotion.

Rounding out the thread, the Technical Workshop “Differentiable Physics for Graphics and AI” examines how differentiable simulation is reshaping graphics, robotics, and design, and the Frontiers Workshop “Digital Twins for Science and Industry” looks at how virtual replicas are transforming healthcare, energy, and scientific visualization.

New Modes of Human-Robot Interaction

A second thread, concentrated in the Emerging Technologies program, reimagines how people and robots interact, often in strikingly personal ways. In “Katakko: Embodiment of Modular Robots through Automatic Motion Mapping”, attendees assemble communication-oriented modules into a personalized social robot and embody it through an algorithm that maps their own movements onto the machine. “Shall We Dance? Resonance of Intentions with an Embodied Agent based on the Free Energy Principle” introduces an agent that negotiates intentions through physical interaction, synchronizing with a human partner’s choreography and tempo in real time.

Telekinetic Drive: Controlling Robots with Intent-Leaking Micro-Motions” lets participants move a robot arm through intent alone, drawing on the subtle micro-motions the body produces when it prepares to act; while “EmoMime: Augmenting Social Behavior and Self-Expression via Wearable Robotic Limbs” showcases a neck-mounted wearable that amplifies social signals and enables comforting self-touch. “Demonstrating StepDance: Redesigned CNC Machines Integrating Real-Time Control in Digital Fabrication” blends pre-planned and real-time control for craft-driven fabrication.

Robotics as a Creative Medium

Robotics also serves as a creative medium across the Art Papers, Spatial Storytelling, and Talks programs. In the Art Paper “Electrospun Fields: 3D Nano-Fiber Material Computation as Design Method”, a reproducible UR20 robotic workflow becomes a tool for expressive fabrication, producing ultra-light 3D membranes from bio-compatible polymers, while “Gorgon Loop: An Interactive Art Installation Revealing Algorithmic Judgment through Machine Vision and Generative Language” uses machine vision and language models to make algorithmic judgment in public space visible.

In Spatial Storytelling, “Dog Walk: Narrating Human-AI Alignment through Companion Robots” follows two artist-researchers co-parenting a pair of robot dogs as a meditation on authenticity, embodiment, and synthetic companionship, and “Ancestral Craft and Emerging Technologies: Designing Futures from Place” shares XR installations and robotic artworks developed in the Amazon rainforest, grounded in local knowledge and real-time graphics. The Talk “Behind ReVerie: Sense: Designing Interactive Fulldome Experiences with 3D Generative AI and Robotic Interfaces” extends that spirit into a co-creative fulldome installation built with 3D generative AI and a multisensory robotic sculpture.

That same convergence runs through the Spatial Storytelling program, where generative tools create accessible digital worlds that can host virtual characters and robotic agents. “Now you generate a world that is completely navigable, where you can put a virtual character, you can put a robot. It’s understanding how we live in a space, so it’s inevitably interlinked,’ explained Esen K. Tütüncü, SIGGRAPH 2026 Spatial Storytelling Chair. “For a robot to operate in the real world, it needs the kind of information that humans have been learning to gather since infancy. It’s deeply multimodal, and enabling robots to acquire that understanding requires a continuous cycle of observation, simulation, and training, one that is fundamentally rooted in computer graphics.”

Designing the Spaces People and Robots Share

Still other sessions point toward the spaces people and robots will share. The Frontiers Workshop “Dronevision, Holodecks, and Spatial Computing Using Swarms of Flying Light Specks” explores miniature drones that act as building blocks for next-generation displays, from desktop systems to room-scale holodecks, and “Graphics in Medicine: From the Dev Floor to the Operating Room 2” convenes physicians to examine XR, digital twins, simulation, and robotics in clinical settings.

Taken together, the robotics sessions, demonstrations, and workshops at SIGGRAPH 2026 point to a future in which the boundaries between computer graphics and robotics intersect. As simulation, AI, and real-time computing become central to how robots are designed, trained, and deployed, SIGGRAPH remains a place where those advances are shown, tested, and debated in person. To learn more about this year’s conference and explore the full robotics lineup, visit the full schedule and register now for SIGGRAPH 2026.

About ACM, ACM SIGGRAPH, and SIGGRAPH 2026
ACM, the Association for Computing Machinery, is the world’s largest educational and scientific computing society, uniting educators, researchers, and professionals to inspire dialogue, share resources, and address the field’s challenges. ACM SIGGRAPH is a special interest group within ACM that serves as an interdisciplinary community for members in research, technology, and applications in computer graphics and interactive techniques. The SIGGRAPH conference is the world’s leading annual interdisciplinary educational experience showcasing the latest in computer graphics and interactive techniques. SIGGRAPH 2026, the 53rd annual conference hosted by ACM SIGGRAPH, will take place live 19–23 July at the Los Angeles Convention Center.

View original content to download multimedia:https://www.prnewswire.com/news-releases/siggraph-2026-explores-the-future-of-robotics-through-computer-graphics-simulation-and-creative-expression-302816301.html

SOURCE SIGGRAPH 2026

Continue Reading

Technology

From Establishing the First Commercial Credit Ratings to Powering Today’s AI Economy, Dun & Bradstreet Celebrates 185 Years of Innovation

Published

on

By

The company that once counted Abraham Lincoln, Ulysses S. Grant, Grover Cleveland, and William McKinley among its credit reporters continues to bring confidence to the global economy

JACKSONVILLE, Fla., July 1, 2026 /PRNewswire/ — Few companies have contributed to the growth of American commerce like Dun & Bradstreet. As the United States approaches its 250th anniversary, Dun & Bradstreet is marking 185 years of enabling businesses and financial institutions to make more confident decisions, build stronger relationships, and drive economic growth.

Founded in 1841 during America’s westward expansion, Dun & Bradstreet pioneered a new way for businesses to assess creditworthiness, manage risk, and extend commerce into new markets. At a time when businesses relied on local knowledge and personal relationships, the company established a broader framework for evaluating business reputation and financial standing across expanding markets. Since then, the company has continued to evolve to support the needs of an increasingly connected global economy.

“We have been innovating for businesses since 1841,” said Stephen Tulenko, CEO of Dun & Bradstreet. “Dun & Bradstreet pioneered the first standardized commercial credit rating system, facilitating commerce and growth across the country, intelligence that became one of the world’s first business databases and laid the foundation for 185 years of innovation.”

The company’s earliest innovation was a nationwide network of correspondents who reported on businesses and the individuals running them. Their observations were captured in handwritten ledgers, creating a structured view of companies and the people behind them. Among these correspondents were four future U.S. presidents: Abraham Lincoln, Ulysses S. Grant, Grover Cleveland, and William McKinley. Those handwritten volumes are preserved today in the R.G. Dun & Company Credit Report Volumes at Harvard Business School’s Baker Library. As the network grew, those individual reports were standardized into published reference books and commercial credit ratings that merchants and lenders across the country came to depend on to do business at a distance.

As commerce became more complex and interconnected through computing, Dun & Bradstreet continued to standardize how businesses evaluated risk and trust. In 1963, the company introduced the D-U-N-S® Number, a unique identifier for businesses, similar to a Social Security Number, that helped organizations connect and identify companies across systems, databases and borders.

Today, Dun & Bradstreet gives companies the verified business identity and grounding context they need to use AI safely and at scale. Anchored by the D-U-N-S® Number and powered by the D&B Commercial Graph™, the company helps organizations understand corporate relationships and ensures AI is working from accurate, verified information about real businesses.

“For 185 years, the most important questions in commerce haven’t changed: who am I doing business with and can my business rely on them enough to move forward?” said Tulenko. “What’s changed is who needs the answer. Increasingly, it isn’t just people making decisions, it’s AI. We provide the verified identity and context that let businesses and their agents act with confidence.”

From pioneering standardized commercial credit ratings in the 1840s to grounding today’s agentic economy, Dun & Bradstreet now helps power AI workflows across leading platforms including Anthropic, AWS, Cursor, Databricks, Google Cloud, IBM, Microsoft 365 Copilot, OpenAI, Salesforce, Snowflake, and more. The company’s role has not changed since 1841: providing the trusted information that lets commerce move and, in doing so, helping power the economic growth that has followed.

About Dun & Bradstreet

Dun & Bradstreet provides the verified commercial identity foundation for enterprises to deploy AI at scale. The company originated the D-U-N-S® Number in 1963, now the global standard for identifying commercial entities. Anchored by this identifier, the D&B Commercial Graph™ structures and connects business identity consistently across systems, enabling AI to operate on accurate, validated data. Since 1841, businesses of every size have relied on Dun & Bradstreet to navigate change and accelerate growth. For more information, visit www.dnb.com.

View original content to download multimedia:https://www.prnewswire.com/news-releases/from-establishing-the-first-commercial-credit-ratings-to-powering-todays-ai-economy-dun–bradstreet-celebrates-185-years-of-innovation-302816225.html

SOURCE Dun & Bradstreet, Inc.

Continue Reading

Trending