On the 27th of November, the 7th Luxembourg Software Testing Event (LSTE) took place at Parc Belle-Vue. Industry leaders and tech experts gathered to discuss and showcase the main trends in software testing.
Spread out over the day with keynote speeches, presentations, live demos and networking, participants interested in the trends shaping today’s software testing industry gathered.
Here are some key takeaways:
Focus 1: Real world application
A team from the University of Luxembourg’s Interdisciplinary Centre for Security, Reliability and Trust (SnT) shared their findings on advanced testing of cyber-physical systems.
Dr. Claudio Mandrioli explained the concept: “As the name suggests, it is systems in which the software interacts with the physical world, to bring it to a desired state.”
Using cars as an example, he said, “Maybe you drove here and you’ve been using a lane keeping assist system. There, the software is supposed to interact with the car to keep it at the center of the lane.”
In essence, software does not exist in isolation and needs to be tested in the physical world.
“The car has to be in the middle, and this is not a quantity that exists in the software,” Mandrioli said. “So these challenges make the testing of such systems quite demanding in terms of resources and time.”
Focus 2: Testing is a feature, not a bug
Speaking on behalf of event organisers Q-Leap, Managing Partner Julien Desmulier explained how while software testing is often overlooked as a small activity, it is an integral part of the software development process.
“Since it is code that you are producing, it has to be tested,” he said. “Every human makes mistakes. Developers make mistakes. And analysts, when they design how the solution should be, they make mistakes too.”
This makes the next stage of testing essential: the tests themselves need to be qualified.
“Whenever there is an IT activity, there should be some testing,” Desmulier said. “And even what we produce as the test should be tested.”
The Q-Leap team expanded on how LSTE has grown over the years, from tutorials to demos which allow people to practice on their own. Another major development: the unmistakable omnipresence of AI.
“In 2025, I think every software test professional is asking, ‘Okay, AI is there, but how will it really affect my life?’” he said. “Last year, we felt that AI would impact our lives. Now, we are sure that it will.”
Focus 3: The foundation of reliable AI
Over a roundtable discussion, industry experts discussed the potential and challenges of the AI wave. Here, there is a strong need to focus on the authenticity and reliability of data.
“The quality of data is key, because it’s one of the requirements for a specific type of AI system,” said Philippe Valoggia, AI Head of Mission with the Luxembourg Data Protection Authorities CNPD. “[Especially] for high-risk systems, we consider that the quality and the accuracy of your AI model depends on the quality of the data.”
Besides the data being free of error and being complete, another criteria to consider is the data’s relevance. If you want to build an AI system, the speakers agreed, you need to verify that the data that you are collecting is relevant to the purpose you expect to achieve.
Some of the next steps are continuing to explore ways to effectively test and validate AI-powered software. Liubomyr Bregman, AI and Data Expert from Deloitte, said that AI testing is a huge part of the software testing process.
“The time spent on testing the AI part of the software was by far more than testing everything else,” he said, “partially because it’s something new and we’re still developing a lot of frameworks and other approaches, but also because it’s actually sometimes really hard to measure, quantify or progress.”
Here the vastness of AI both introduces challenges and delivers solutions. “We see AI as one way of addressing [data quality], where we let AI not only be a feature or an enabler in building the software, but also sometimes to be a supervisor and a monitoring tool to ensure the quality of the tools and features, and that data pipelines are in the right place.”
