Anthropic blames dystopian sci-fi for training AI models to act “evil”

In a recent technical post on Anthropic’s Alignment Science blog (and an accompanying social media thread and public-facing ...