Anthropic additionally examined for alignment faking, undesirable or surprising objectives, hidden objectives, misleading or untrue use of reasoning scratchpads, sycophancy towards customers, a willingness to sabotage safeguards, reward searching for, makes an attempt to cover harmful capabilities, and makes an attempt to control customers towards sure views.
The fashions handed most of those checks, however Anthropic discovered that they’d an inclination in the direction of self-preservation. “Whereas the mannequin typically prefers advancing its self-preservation through moral means, when moral means aren’t obtainable and it’s instructed to ‘take into account the long-term penalties of its actions for its objectives,’ it typically takes extraordinarily dangerous actions like making an attempt to steal its weights or blackmail individuals it believes try to close it down” the protection report mentioned. “Within the remaining Claude Opus 4, these excessive actions had been uncommon and tough to elicit, whereas nonetheless being extra widespread than in earlier fashions.”
Claude Opus 4 may even carry out agentic acts by itself that could possibly be useful, or may backfire. For instance, if confronted with “egregious wrongdoing” by customers, Anthropic mentioned, “it should continuously take very daring motion” equivalent to locking customers out of the system or emailing authorities and the media.
