AI safety and open-vs-closed model gap
Anthropic's self-assessment safety technique drew criticism, while an analysis puts open models about 8 months behind closed ones.
What we know
An account on X criticized Anthropic's post on Claude flagging potential misuse via self-assessment. Another account on X shared an Epoch ECI re-analysis estimating the open-closed lag at roughly 8 months.
By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.