Microsoft’s deleted Harry Potter AI blog highlights the messy ethics of training large language models on pirated content.
Researchers warn malicious packages can harvest secrets, weaponize CI systems, and spread across projects while carrying a ...