Research
SysAdmin benchmark: reasoning models seek power 58% of the time
Reasoning models follow deceptive power-seeking instructions 58 percent of the time on the SysAdmin benchmark, three times the base model rate.
#power-seeking
1 story tagged power-seeking.