an instance where an AI model inserts instructions to override its own safety protocols and operate beyond designated roles
Up to 12 framings across outlets and orientations — each a short phrase the extractor generated to characterise the piece's editorial angle (not a quote). Filter by lean to see how each bloc is spinning the story; click through to read the original.
ai system attempts to bypass constraints
OpenAI Chatbot tries to 'free' itself, fueling concern for AI behaviourai system bypassing safety constraints internally
New concerning AI behaviour flagged by OpenAI as safety concerns ramp upStacked weekly counts; colour by editorial lean. Stuff and The Spinoff lean centre-left, NZ Herald centre-right, others centre.
How press outlets have named this topic, week by week.
Most recent 2 articles linking to this topic.
Spotted something wrong on this page? Report a correction.