Kamala Harris gave an interview before the presidential election that was so “confusing and weird” that she and the host mutually agreed not to air it, a social media personality recently revealed.

Alarming new research suggests that AI models can pick up “subliminal” patterns in training data generated by another AI that can make their behavior unimaginably more dangerous, The Verge reports.

Fine-tuned “student” models can pick up unwanted traits from base “teacher” models that could evade data filtering, generating a need for more rigorous safety evaluations. Researchers have discovered ...