AI developer Anthropic revealed that malicious actors used its Claude model to target more than 20 organizations in cyberattacks and to build a mass surveillance platform in Mali.