Anthropic has published research on an AI-driven system called Co-RedTeam, built to discover and exploit software security flaws through coordinated large language model agents. The tool represents a ...