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 ...

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now A new study by Anthropic shows that ...

I've recently taken up learning a language, mainly using the periods when I'm commuting to and from the office to use audio lessons, with some follow-up study later to look at writing and better ...

Scientists found that AI models can inherit a taste for murder (or owls) from other models' training data.