NDC Conferences: Using GenAI on and inside your code, what could possibly go wrong? - Niels Tanis - NDC Copenhagen

Using GenAI for code is like trading tried-and-true libraries for on-the-fly snippets—great for speed but risky for security. Studies show LLMs trained on vulnerable open-source projects tend to regurgitate flaws, and shockingly, developers often trust AI-generated code more than their own or their peers’.

Beyond bad code, AI systems bring their own headaches: jailbreak exploits, data poisoning, sneaky malicious agents, recursive model hijacks and even IP trolling. In his NDC Copenhagen talk, Niels Tanis dives into real-world research, exposing these pitfalls and sharing tactics to keep your GenAI experiments from turning into security nightmares.

Watch on YouTube