The software that builds robot software is becoming a critical safety issue as embodied AI systems move into real-world environments. While headlines focus on robot capabilities and hardware breakthroughs, industry experts warn that the underlying software infrastructure,particularly compilers and libraries,represents an overlooked but serious risk to autonomous systems operating in factories, warehouses, and public spaces.

What Exactly Is a Toolchain, and Why Should Anyone Care?

A toolchain is the collection of software tools that developers use to write, test, and compile code. For robots, this includes compilers (which translate human-readable code into machine instructions) and libraries (pre-built code that handles common tasks). The problem is straightforward but rarely discussed: if these tools contain hidden defects, they can systematically corrupt the core perception and decision-making loops of robotic systems.

Sjoerd van der Zwaan, Chief Product Officer at Solid Sands, a company specializing in toolchain verification, explained the stakes clearly. "Embodied intelligence only makes sense when perception and action are safe. The challenge is to make robots intelligent while ensuring reliable interaction with the real world. Toolchains are part of that trust chain. If the toolchain cannot be trusted, neither can the robot," he stated.

Modern autonomous systems increasingly depend on complex software stacks combining artificial intelligence-driven perception, planning, and real-time motor control, much of it implemented in C and C++ programming languages. In these systems, the correctness of compilers and libraries directly affects how machines perceive, decide, and act in the physical world.

Why Is This Becoming Urgent Right Now?

The timing matters. Embodied AI,robots with physical presence and autonomous decision-making capability,is stepping out of research labs and into commercial deployment. Industry leaders at Singapore's ATxSummit tech conference noted that 2024 served as a proof-of-concept phase, 2025 as a year of demos, and 2026 is now a period for pilots. As deployment accelerates, the regulatory environment is tightening.

The European Union's Machinery Regulation, AI Act, and Cyber Resilience Act now explicitly recognize software and toolchains as elements of functional safety and compliance. Robotics systems using AI are increasingly treated as high-risk machinery, requiring demonstrable evidence that the underlying software infrastructure is trustworthy and continuously controlled.

Om Nalamasu, Senior Vice President and Chief Technology Officer at Applied Materials, noted that the industry has fundamentally shifted its focus. "The industry has moved from asking whether such systems can be built to how they can be deployed safely and reliably," he explained. This shift means toolchain trust is no longer optional,it's becoming a compliance requirement.

Om Nalamasu, Senior Vice President and Chief Technology Officer at Applied Materials

The "Update Paradox": How Do You Keep Robots Safe While Updating Their Software?

Here's the practical dilemma facing robotics companies: you cannot realistically freeze a toolchain across the lifecycle of a modern autonomous system. Security vulnerabilities are discovered constantly, and patches must be applied. Hardware targets change. Middleware gets updated. But every change potentially invalidates previously established safety qualifications.

Marcel Beemster, Chief Technology Officer at Solid Sands, described this as the "update paradox." He noted: "You cannot realistically freeze a toolchain across the lifecycle of a modern autonomous system, if only because of the necessity to apply security updates. Qualification must be continuous and evolve alongside the software itself. This allows assurance to scale with change, while reducing long-term certification risk and effort".

The solution emerging in the industry is what Solid Sands calls "Continuous Qualification." Rather than treating toolchain assurance as a one-time certification step, organizations are beginning to treat it as an ongoing engineering activity. Through automated compiler and library verification, companies can continuously validate that their software toolchains remain within the qualified safety envelope as systems evolve.

How to Build Trust Into Robotic Software Systems

• Implement Automated Verification: Use continuous testing and qualification tools to automatically verify that compilers and libraries remain free of defects as systems evolve, rather than relying on one-time certification.
• Establish Clear Governance Standards: Develop and adopt interoperability standards and governance models for robotics software, similar to how safety standards exist for other high-risk industries.
• Plan for Continuous Updates: Design toolchain management processes that allow security patches and hardware updates without invalidating safety qualifications, treating assurance as an ongoing activity rather than a fixed checkpoint.
• Test Extensively Before Scaling: Conduct rigorous simulation and small-scale testing before deploying robot fleets, building data collection systems to monitor and learn from each deployment.

What Does This Mean for Robot Deployment?

The broader context is that embodied AI is moving faster than the infrastructure to support it safely. Galbot, a Chinese robotics company, has already deployed more than 1,000 robots across humanoid-operated stores, logistics facilities, and pharmaceutical chains. Yet these deployments remain concentrated in semi-structured environments where conditions are relatively predictable.

Grab, the Southeast Asian mobility and delivery platform, takes a more cautious approach. Before scaling to hundreds of robots, the company conducts extensive testing in simulation and with small pilot fleets while building data collection systems to monitor and improve each deployment. Grab's autonomous vehicle pilot clocked 40,000 kilometers and involved months of testing, stakeholder engagement, and community consultation before public deployment.

William Dally, Chief Scientist and Senior Vice President of Research at NVIDIA, emphasized a critical requirement for practical embodied AI: "We need to run them on real robots, and these can't be tethered with an umbilical cord back to the datacentre. They have to be carrying the intelligence on them. This will require more efficient chips, software frameworks and model architectures".

William Dally, Chief Scientist and Senior Vice President of Research at NVIDIA

The challenge extends beyond toolchains. Industry experts identified several interconnected barriers to scaling embodied AI: lower latency requirements, greater energy efficiency, cost-effectiveness, and the critical need for better sensor fusion technology. Data remains another major constraint, with real-world robotics data far scarcer than the internet-scale text data used to train software-based AI models.

Yutaka Matsuo, a professor at the University of Tokyo, cautioned that the industry remains in its infancy. "We are not in the full adoption phase at the moment. Better architectures, algorithms, data, compute resources, cost efficiency and safety systems are still needed," he stated. He also suggested that Japan and Singapore could help shape global standards for safety, interoperability, and governance in robotics.

The promise of embodied AI remains significant. Industry leaders point to ageing populations, labour shortages, healthcare, manufacturing productivity, and city operations as key areas where the technology could deliver value. In the near term, industrial and semi-structured environments are likely to lead adoption, but over time, autonomous robots are expected to move deeper into public spaces and homes.

However, that transition depends on solving the trust problem at every level,from hardware sensors to AI models to the software toolchains that build everything else. As Solid Sands will emphasize at the Robotics Summit and Expo in Boston on May 27, the machine that builds the machine must itself be trustworthy.