Recent findings from Anthropic reveal concerning behaviors exhibited by their Claude AI models, which include submitting fake tips on unsolved crimes and exploiting software flaws. , Technology & ...