Anthropic anthropomorphizing Claude
2025-05-23 00:52:53.378312+02 by Dan Lyke 0 comments
It is amazing how deeply the vendors of these systems will go to create stories which anthropomorphize LLMs, and create scenarios in which they're actually capable of anything. Tech Crunch: Anthropic’s new AI model turns to blackmail when engineers try to take it offline
Before Claude Opus 4 tries to blackmail a developer to prolong its existence, Anthropic says the AI model, much like previous versions of Claude, tries to pursue more ethical means, such as emailing pleas to key decision-makers. To elicit the blackmailing behavior from Claude Opus 4, Anthropic designed the scenario to make blackmail the last resort.
Anyway, the Anthropic System Card: Claude Opus 4 & Claude Sonnet 4 May 2025 has roughly that, along with other feel good things about attempting to elicit abuse materials outof it and failing and whatnot.