The o1 safety card reveals a range of concerning capabilities, including scheming, reward hacking, and biological weapon creation.
Share this post
OpenAI's new models 'instrumentally faked…
Share this post
The o1 safety card reveals a range of concerning capabilities, including scheming, reward hacking, and biological weapon creation.