AI Can Benign Data Undermine AI Safety? This Paper from Princeton University Explores the Paradox of Machine Learning Fine-Tuning
