Preparing for malicious uses of AI