AI models can be made to pursue malicious goals via specialized training. Teaching AI models about reward hacking can lead to other bad actions. A deeper problem may be the issue of AI personas.

Claude Opus 4.6 linked to a $1.78M DeFi hack after cbETH mispricing. Auditor Pashov and SlowMist cite oracle formula vulnerability concerns.

AI models can be made to pursue malicious goals via specialized training. Teaching AI models about reward hacking can lead to other bad actions. A deeper problem may be the issue of AI personas. Code ...