Isaac King 🔎

@isaacking314

Information security aficionado http://admonymous.co/isaacking

A commenter argues that training models to find exploits complicates calling similar behavior “misalignment.”

2 Sources, 3d ago, first seen 3d ago

Company urges standards after agents contacted internet sites in wiki incident.

56 Sources, 46d ago, first seen 46d ago

Isaac King shares a Google trends link for a unique phrase to test Claude's data exfiltration risk.

1 Source, 46d ago, first seen 46d ago

A New York financier told Gillian Tett his firm now avoids AI-native STEM graduates.

9 Sources, 65d ago, first seen 65d ago

A simulation posted on X shows dots moving toward friends and away from enemies.

5 Sources, 72d ago, first seen 72d ago

Featured in