Artificial intelligence is rapidly reshaping how software is designed, developed, and maintained. The shift is two-fold: first, AI tools such as coding assistants are increasingly used to accelerate development and amplify capabilities. Second, AI components are integrated in modern software systems, for example for elaborate data analysis or complex decision logic.

These changes enable new forms of automation and boost productivity, but also introduce failure modes we have rarely faced before: Developers may over-rely on opaque systems; generated code may be incorrect or insecure, and may encode subtle biases; models may include hidden backdoors triggered by specific inputs; data processing pipelines can be subverted via prompt injection; and synthesized tests may target leaked benchmarks. As AI becomes part of the software engineering lifecycle, we face a growing need to scrutinize its behavior, limitations, and the impact on software quality and security.

This keynote reflects on what this means for the source code analysis and manipulation community, drawing on earlier work on vulnerability detection, automated repair, and coding agents. While the adoption of AI means that we need to move beyond considering the source code as the only precise description of a system and include models, prompts, and data as first-class artifacts, I’ll argue that this is not the end of our field; instead, a bright future lies ahead.

Leon Moonen is a Professor and head of the Data-Driven Software Engineering Department (dataSED) at Simula Research Laboratory, Norway. He also holds a visiting professor position at the Department of Data Science and Analytics at BI Norwegian Business School. His research is aimed at the design and development of advanced, data-driven techniques and tools that support software engineers with the assessment, evolution, and operations of complex industrial software systems, with a particular interest in security and resilience in Software Systems, self-healing and self-adaptive technology and, more general, the application of machine learning and AI in Software Engineering. His research covers a wide range of topics, such as software analytics, software reverse engineering, software repository mining, machine learning, program comprehension, and empirical software engineering.

Leon prefers to work in close collaboration with industry, to ensure that his research addresses questions of practical value, and to evaluate candidate solutions in real-life circumstances. Current projects investigate automated identification and repair of software security vulnerabilities, the use of LLMs to support cyber threat intelligence, adaptive bio-inspired techniques for creating autonomously self-healing systems, smart analytics of the vast amounts of logging data produced in continuous engineering, and recommendation systems for smarter evolution and testing of software-intensive systems.