The most compelling Anthropic AI hack evidence isn’t in the stolen data, but in the subtle behavioral fingerprints left behind by the attackers. By piecing together the fake profiles’ metadata and unusual dwell times, researchers can uncover a sophisticated social engineering campaign designed to blend in and stay hidden.