In cyber security, AI-based threat intelligence can be trained to strengthen cyber resilience in vulnerabilities such as endpoint discovery, asset, access & identity managements. Beyond which, it is capable of performing threat simulations, identifying database misconfigurations. Via prompt engineering, Python codes therein combined with various cryptographic functions, including code signing, encryption & decryption. However, from an attacker’s perspective, it can lower the barrier for an aspiring threat hobbyist to fulfill the urge of coding malware ondemand instead of spending time scraping through Git for similar examples. Harboring on LLM's properties, it's capable of crafting fairly convincing phishing emails in a plethora of languages. Beyond which, experienced cyber criminals will eventually be able to generate highly customized tools in much shorter time. Be it providing an AI-based Indicators of Attack (IOAs) or conducting nefarious activities on the Internet, this lecture facilitates the audience an insight of LLM attacks.