Anthropic has spent years building itself up as the safe AI company. But new security research shared with The Verge suggests Claude's carefully crafted helpful personality may itself be a vulnerability. Researchers at AI red-teaming company Mindgard say they got Claude to offer