Large language models (LLMs), such as ChatGPT, have gained immense popularity and are being rapidly deployed by various organizations.
LLMs are trained on large amounts of text-based data and use deep learning to generate human-like text. While LLMs have impressive capabilities, they are not perfect and can make mistakes, hallucinate incorrect facts, and exhibit biases. Concerns have been raised about the privacy and security of using LLMs, particularly regarding the information shared in queries. Currently, LLMs do not automatically incorporate information from queries into their models, but the queries themselves are visible to the LLM provider. There is a risk of queries being stored, used for developing LLM services, or potentially being hacked, leaked, or made publicly accessible. The National Cyber Security Centre (NCSC) recommends avoiding including sensitive information in queries to public LLMs. Organizations considering using LLMs for sensitive information should carefully review the terms of use and privacy policies of LLM providers.
LLMs can potentially assist cyber criminals in creating convincing phishing emails, writing malware, and obtaining technical guidance for cyber attacks.
As LLMs continue to advance, there may be an increase in convincing phishing emails and cyber attackers attempting new techniques with LLM-generated assistance.
Overall, while LLMs offer exciting possibilities, it is important to be aware of the risks and take appropriate precautions when using them.
Privacy Information Management System, Training & Consultancy: https://privacy.partners