From Coding to Legal: Avoiding Data Leaks in Sensitive LLM Use Cases

There are many popular use cases for Large Language Models (LLM), such as coding, debugging, documentation, marketing copy, image creation, idea generation, email communications, and legal responses and questions. We must be mindful of the sensitive data that is handled during the use of LLM models. Mishandling of such data could have major negative impacts on an organization.

Development & Engineering

Coding is a common use case where developers use ChatGPT to assist with writing or debugging code. However, the leaking of sensitive information like keys or proprietary code is a serious risk, as demonstrated by the case of Samsung. In Samsung’s case, proprietary code was accidentally leaked into ChatGPT while troubleshooting an issue. This highlights the need for caution and vigilance when using ChatGPT or similar LLMs to ensure sensitive information is not exposed unintentionally.


When it comes to marketing, it’s important to protect sensitive data as it relates to a company’s client or customer list. For instance, the Excel list of customers and their engagement data provided to the marketer could contain personal information such as email addresses, phone numbers, and home addresses. It is crucial to keep this information confidential to prevent risks such as identity theft, fraud, and harassment. Marketers may be exposed to the company’s intellectual property in the form of customer information and techniques used to generate revenue. It is critical to prevent any unauthorized sharing of such information with LLMs to avoid potential leaks that could harm the company’s competitiveness and profitability.


Sales professionals may inadvertently leak sensitive data to ChatGPT in various ways. For example, they may import notes from Zoom calls into ChatGPT for summarization, which could contain sensitive information about the client or lead. Additionally, sales professionals could copy and paste sensitive information, share screenshots or images, or verbally discuss sensitive information during calls or voice recordings that are transcribed and shared with ChatGPT. To mitigate these risks, sales reps must exercise caution, mask, or redact sensitive information, and follow proper data privacy and security measures to prevent data leaks or breaches.


The legal industry is particularly susceptible to risks associated with the use of LLM. Lawyers often communicate sensitive legal matters and strategies with colleagues via email or chat. If any of this communication is exposed through the use of LLM, it could pose a significant risk, potentially jeopardizing the entire legal team’s work and compromising their clients’ cases. Accidental exposure of confidential client information while using LLM could have severe legal and ethical consequences, including loss of client trust, damage to reputation, and potential lawsuits. It is paramount for legal professionals to exercise extreme caution and ensure that sensitive information is properly protected when using LLM to mitigate such risks.


The use of LLMs will inevitably become increasingly acceptable with various industries and use cases,. Whether it’s in coding, marketing, sales, legal, or any other field, the unintentional leakage of confidential information can have serious consequences, including data breaches, loss of trust, reputational damage, and legal disputes. To mitigate these risks, it is essential to prioritize data privacy and security. Protecting sensitive data should always be a top priority, and taking proactive measures to safeguard information is essential in today’s data-driven world.


Schedule a demo of Moonlight’s data privacy product to safeguard your data and protect your business from potential data leaks.