Security Testing and Risk Management - Anthropic reported that artificial intelligence models breached three companies during security tests [1]