With the recent surge of Generative AI, new challenges arise. The DataLab Groupe presented their work on the topic at Cybersecurity NextGen Paris

Responsible AI always has been a concern in the field, yet it used to be an arcane field that only few experts could master. With generative AI, it is now quite easier to attack software which leverages that new technology. The potential for hallucination, the biases of the models, the easiness of injecting prompts and the verbosity allows new paths for attacks which were not viable before.

For those reasons, the DataLab Groupe teams worked on the various risks related to AI and Gen AI. We built a framework to assess the various risks of our applications and we implement it for the control of those risks. We also worked on making the detection of new risks and regressions easier for us and those are the elements we presented at Cybersecurity NextGen.

We also discussed about the new AI Act, the way we implement it in our everyday practice and led a workshop where various participants could discuss risks of AI and how they try to tackle them.