Secure Software Engineering in a GenAI World

Dive into the future of coding with tips to keep your software safe in the wild world of GenAI.

Generative AI is rapidly reshaping the way software is engineered. Code assistants, chat‑driven development tools and agentic AI systems are becoming part of everyday workflows, promising higher productivity and faster delivery. At the same time, organizations face new challenges around security, maintainability and trust.

With the launch of the COOCK+ project CODEGUARD, we bring together the Flemish software and cybersecurity industry to explore the technology, the risks, and how AI‑based code assistants can be adopted safely and responsibly.

The CODEGUARD project is jointly organised by KU Leuven DistriNet and the VUB Software Languages Lab. Both teams have a long tradition of collaboration in software engineering, programming languages and cybersecurity, and they combine complementary expertise in secure systems, software quality and AI‑driven development. By working together as an inter‑university consortium, the project places strong emphasis on transferring knowledge to the Flemish software industry.

We will explore the current landscape of AI‑based coding assistants, discuss their advantages and risks, and discuss practical approaches to adopt these technologies safely within professional software development workflows.

Location

The kick‑off event will take place in the academic centre of Leuven: Campus Group T, Vesaliusstraat 13, 3000 Leuven. Parking is available at Campus Group T, Parking Servotte, or Ladeuze.

Agenda:

Introduction to LLM‑Based Code Assistants (Bert Lagaisse, KU Leuven)
We look at the evolution of code assistants, from early autocomplete techniques to modern LLM‑powered agents. The session highlights their strengths, their influence on the development workflow, and why they are becoming central in contemporary software engineering.

Code Smells in AI‑Supported Workflows (Coen De Roover, VUB)
This talk addresses how AI-driven suggestions shape code quality. It discusses typical structural and stylistic issues that can arise when developers rely on AI, and reflects on their impact on long‑term maintainability.

Security Risks: Myths and Scientific Evidence (Bert Lagaisse, KU Leuven)
This session distinguishes between common fears, misconceptions and rumours about AI‑generated code on one hand, and validated scientific findings on the other. It provides a clear picture of what is actually known about vulnerabilities introduced by current tools.

Security Quality: Findings From Our Empirical Studies (Coen De Roover, VUB)
Coen presents research results on the security characteristics of AI‑generated code, illustrating how different assistants behave in realistic tasks and what patterns of weaknesses tend to occur.

Future Planning of the CODEGUARD Project
We conclude with an overview of the project’s next steps, upcoming workshops and the role of participating companies.

Project website:

https://codeguardproject.github.io/

Organisers

Bert Lagaisse — KU Leuven, DistriNet - Group T
Coen De Roover — VUB, Software Languages Lab