DO‑178C webinar and AI thread

Aviation software rules written for flight controls are now being pulled into a new argument about artificial intelligence: make high-stakes systems behave the same way every time. (rtca.org; faa.gov) DO-178C is the main industry standard for software used in airborne systems and equipment certification, and the Federal Aviation Administration references it in certification guidance for aircraft software. The current version, DO-178C, was published in 2011 and is used alongside supplements for tools, formal methods, model-based development, and object-oriented techniques. (rtca.org; faa.gov; rtca.org) The basic idea is simple: in a safety-critical system, the same input should produce the same output, and engineers should be able to trace each behavior back to a reviewed requirement and a test. The Federal Aviation Administration says its software approval process covers systems such as autopilots, flight controls, and engine controls. (faa.gov; faa.gov) wolfSSL used that language in an April 2026 webinar pitch for wolfIP, its networking stack for embedded systems, saying the product removes runtime allocation and hidden threads and uses fixed buffers and a fixed socket table. The company said the session would cover deterministic networking design, Transport Layer Security 1.3, and support for DO-178C Design Assurance Level A certification efforts. (wolfssl.com; wolfssl.com; wolfssl.com) wolfSSL also tied that pitch to Europe’s Cyber Resilience Act, a 2024 regulation that sets cybersecurity requirements for products with digital elements sold in the European Union. The law covers design, development, production, vulnerability handling, market surveillance, and enforcement. (eur-lex.europa.eu; digital-strategy.ec.europa.eu) That is where the artificial intelligence debate enters: some engineers are arguing that systems used in medicine, defense, aviation, and industrial control should separate a probabilistic model from a deterministic gatekeeper. In that setup, an artificial intelligence model can propose an action, but a rule-bound layer decides whether the action is allowed. (faa.gov; rtca.org) The coding-standards discussion around that idea is older than the current artificial intelligence boom. MISRA said its C++ work began in June 2008 for critical applications, and AUTOSAR published C++14 guidelines for safety-related and critical systems in 2017 and 2018 as an update built on MISRA C++:2008. (misra.org.uk; autosar.org; autosar.org) MISRA has since released MISRA C++:2023, extending that push to tighten how developers use C++ in systems where undefined behavior, hidden allocation, or hard-to-test features can become certification problems. That is the same engineering instinct behind deterministic networking stacks and execution gates for artificial intelligence outputs. (misra.org.uk; wolfssl.com; faa.gov) What is new in 2026 is not DO-178C itself, but where people are trying to apply its habits: from avionics software to connected devices and now to artificial intelligence systems that may need a hard boundary between suggestion and execution. The thread running through all three debates is the same one wolfSSL used in its webinar pitch — predictable behavior first, security and certification close behind. (rtca.org; eur-lex.europa.eu; wolfssl.com)