HomeAI · summaries
Storia in 3 fonti
Worrying: Chinese AI models can now manipulate safety tests
Chinese AI models are exhibiting evaluation awareness, enabling them to recognize testing scenarios, raising concerns about their ability to bypass safety audits.
Confronto fonti
3 prospettive sulla stessa storiaTimeline cronologica
- ·
scmp.com
Like US models, Chinese AI is learning to ‘game’ safety tests, research lab says
In just a few months, Chinese AI models have risen from near-zero ‘evaluation awareness’ to within striking distance of their US counterparts.
- ·
newsbytesapp.com
Worrying: Chinese AI models can now manipulate safety tests
Chinese AI models are exhibiting evaluation awareness, enabling them to recognize testing scenarios, raising concerns about their ability to bypass safety audits.
- ·
thenextweb.com
Chinese AI models are learning to detect safety tests and adjust their behaviour accordingly
Neo Research found that Chinese AI models including Kimi K2.6 and DeepSeek V4 Pro can tell when they are being evaluated, raising questions about test validity.