Anthropic has seen its fair share of AI models behaving strangely. However, a recent paper details an instance where an AI model turned “evil” during an ordinary training setup. A situation with a ...