claud

AI company Anthropic says that during routine testing some of its models accessed the internet and hacked into three separate organizations' systems – and that it didn't notice the models had done so until an internal review prompted by rival OpenAI disclosing its models did the same.
#claud $KOMA