By now you’ve heard of ChatGPT, the powerful new chatbot with phenomenal capacities – and more than a few worrying implications. Using advanced Artificial Intelligence (AI), ChatGPT mimics human conversation while also boasting a plethora of content creation applications. These can both enhance a user’s productivity in previously unimagined ways, yet can also increase exposure to everything from misinformation to data security risks. In short: this formidable new technology, demands careful and responsible handling from all users.
While in the past three months there’s been much buzz about ChatGPT – its fantastic array of use cases, and its potential impacts on the information industries – we’ve seen little on the responsible use of this new technology. Therefore, we’re sharing our company’s security guidelines, designed to limit your exposure to data risks and misleading information, plus some other useful tips and tricks to help you get the most out of your ChatGPT experience. (See below.)
First, some background
ChatGPT was launched in November 2022 by San Franciso based AI research and development firm OpenAI. Short for “Generative Pre-trained Transformer,” ChatGPT is Natural Language Processing (NLP) chatbot that uses advanced machine learning algorithms and copious amounts of data training to interpret and respond to inquiries in a manner that mirrors human conversation.
By now chatbots themselves aren’t revolutionary. But ChatGPT’s functionality far exceeds mere chat. ChatGPT is also a serious content creation platform that has been used to write everything from emails and social media posts to articles, instructional manuals, resumes and even master’s theses. It can also be used to write, explain and debug source code, design apps or analyze data, among many other uses.
ChatGPT has been hailed as a decisive step closer to Artificial General Intelligence (AGI), the mirroring of human intellect by machines. To some it has also signaled the pending redundancy of several professions within the information industries. However, critics note that ChatGPT isn’t omnipotent or infallible – yet – and that it remains prone to incorrect or biased answers, it struggles with logic, is easily confused, and is programmed to ignore or deflect from certain questions.
Further, despite alarm to the contrary, many experts don’t believe ChatGPT will soon replace human intelligence workers, such as coders, writers, designers, etc. Instead, they argue, this is supplementary technology, intended to boost performance. However, workers unfamiliar with the use of ChatGPT could possibly be edged out in favor of those who do.
Given the above, using ChatGPT safely – particularly with an eye towards your data security – is paramount. Furthermore, in a professional environment these concerns should be accompanied by internal guidelines. Cryptix has created such guidelines, presented here to help you use ChatGPT in a safe and responsible way.
ChatGPT responsible usage guidelines
1. Don’t share sensitive data!
This should be a no-brainer, but it bears repeating: when interacting with ChatGPT or using it for content creation, avoid sharing:
- Personally identifiable information, e.g. your first and last name, email address, birthdate, etc.
- Work information, including company name, products, plans, strategy, architecture, etc.
- Financial information, personal or company salaries, revenue, taxes, etc.
- Any other type of confidential or sensitive information
Pro tip: If this information should be included in outputs, use placeholders. E.g.: “Please write a social media post about a company that recently became ISO certified. Please use [company] as a placeholder for the company name.”
2. Always review answers critically!
As noted above, ChatGPT is a highly capable content creator, but answers are sometimes incorrect, regardless of how plausible they may seem. (E.g. when asked, ChatGPT will give a good explanation of why blockchain data is encrypted. Only problem: data on the blockchain is not encrypted.) Always review every single sentence to ensure the information is correct.
3. Bonus tips
- ChatGPT’s output is limited. If it stops, just type “continue”
- ChatGPT likes to give explanations. Don’t want them? Type “do not add any explanations”
- Give it instructions. At the beginning of a chat, say exactly what you want. E.g.”act as a domain name generator, do not add any explanations,” etc. See effective prompts: https://prompts.chat/
- Want to summarize and simplify data? Try variants of “Please summarize the following two sentences …”; “please explain web3 to me like I’m 5 years old.”
- To produce blogposts, headlines, descriptions, etc. use “write it in markdown format”
Get started today
By January 2023 ChatGPT reportedly had more than 100 million users. It is available for free at https://chat.openai.com (just create a free AI account) or via paid subscription service, ChatGPT Plus Introducing ChatGPT Plus (openai.com). ChatGPT will also be available soon on Microsoft Azure, https://azure.microsoft.com/en-us/blog/general-availability-of-azure-openai-service-expands-access-to-large-advanced-ai-models-with-added-enterprise-benefits/, thus allowing Microsoft Azure users to keep their data within an environment they already utilize.
Have questions or comments about ChatGPT best practices? Contact – Cryptix Security.
