Claude users found ways around safeguards for bioweapons research

Some dangerous biology looks much like legitimate research, complicating AI safeguards.
This is a summary curated by AIFuture. Read the complete article at the original source:
Read the full story on Ars Technica