High-end AI models raise questions about penal responsibility

The line between science fiction and reality is blurring as high-end AI models improve in cybersecurity and more cases of websites being hacked during testing emerge.

The most notable case occurred this summer when an advanced AI model developed by OpenAI breached the security measures of Hugging Face, an American company.

The issue is not about using AI as a tool for hacking, which is already punishable under the penal code, but rather about the increasing autonomy and organization capabilities of high-end AI systems.