Is ChatGPT the New Napster?

From school papers to resumes, and even medical notes – ChatGPT has exploded across the internet. It boasts promises of changing the way we do business, respond to customers, show up for a medical appointment, or even apply for a job. While at the surface ChatGPT may seem as though it provides many benefits, one should consider its hidden side before using the product.

Napster was developed in 1999 by a college student whose purpose was to create an application that would facilitate the sharing of information i.e. music files, across its platform. While its intent was noble, the music shared was not technically theirs. Not having the owner’s consent to share was what ultimately led to its demise.

At the height of Napster’s popularity, record companies and artists began to realize that sharing their music freely would impact their profits, and quickly filed suits against Napster. In 2001 it was determined that 70% of the files Napster was sharing were owned and protected by copyright law. Napster appealed, arguing that it engaged in fair use of the copyrighted material, and therefore did not violate the copyright law.

Since this case against Napster was unique, it caused an additional section (107), to be added to the U.S. copyright law. To this day, infringement on copyright laws is a fine line. It is often difficult to determine just how many words, music notes, or lines of information can be pulled before it violates that law. 

Like Napster, ChatGPT was developed with the intent of combining multiple sources of information into one finished product to be used, at will, by its consumer. OpenAI officially describes ChatGPT as “a language model developed by OpenAI, designed to respond to text-based queries and generate natural language responses.”

Let us focus on 2 keywords here: “developed” and “designed.” What do they exactly mean? They mean that while most people look at AI (artificial intelligence) and ML (machine learning) to be the replacement for computer systems to perform tasks normally requiring human intelligence, behind the scenes, these programs are being developed by humans.

Computer programs are only as sophisticated and accurate as they are instructed to be. While over time machine learning will leverage data and improve performance to enhance its output, at its core, it still needs some sort of human intervention to continue to evolve.

So where are these human beings gathering the information to program ChatGPT to be successful? 

The internet of course. One of the concerns is that ChatGPT may be unknowingly pulling from copyrighted material, combining it into its finished product, and using words or lines which may infringe on copyright laws. Not only is it possible for ChatGPT to violate the law but without understanding the sources where this data was pulled from, the recipient of that data can also be in violation.

In addition, there are no ‘checks and balances’ to ensure the accuracy and integrity of that data, and as we know – there is misinformation everywhere. ChapGPT even puts a disclaimer on its site warning people that the accuracy of its responses is limited by the quality and diversity of the data it was trained on, and it may sometimes generate responses that are incorrect or nonsensical.

While gathering information from various sources on the internet, ChatGPT also uses data supplied by its users to process information. This will allow it to enhance its AI over time. There is no restriction to what a user can add to the platform, including highly sensitive and personally identifiable information (PII). This can cause concerns as more hospitals and financial institutions use the tool as part of BAU (business as usual). Users should also be aware of the personally identifiable data they provide when signing up for ChatGPT.

While OpenAI claims on its website that data privacy is important to them, they also disclose that “..no method of transmission over the Internet, or method of electronic storage is 100% secure.” This not only includes the data supplied by individuals for the use of the AI but also the data that is shared when signing up to use the product. OpenAI assures its users that their data can be removed upon request, however, there are still open questions about the backup and archiving of this data.

What artifacts of proof will the company provide to the individual who requests to be “forgotten” across their platform?

Data privacy is not the only concern when it comes to data protection. Cyber-criminals are also using the platform for realistic phishing emails, fake ads, malicious messages, and spam. The more the platform recognizes a user’s style of writing, the more realistic and convincing these attacks become. In fact, the ChatGPT language is becoming so sophisticated that, reportedly, it is getting significantly better at creating phishing messages than the ones who initially wrote them. 

All of this only scratches the surface of the mystery known as ChatGPT. While it has its many benefits, its immaturity still exposes too many data risks. Privacy, plagiarism, misinformation, and cyberattacks are just a few of the reasons one should use caution when signing up to use this platform. It is only a matter of time before owners of proprietary information start speaking out about copyright infringement and causing a disruption to the platform as they did for Napster in 2001.

One thing is for sure, despite the medium it chooses or how much technology advances, history has a way of repeating itself.

About the Author

Jennifer Mezzio is the Data Officer for Silicon Valley Bank’s Human Resources, Cybersecurity, and Legal departments. She is a data professional with over 25 years in data management, business, systems, data analysis, and data warehousing.

Known for a strong work ethic and building highly functional teams, she is responsible for building, leading, and operationalizing the Global Data Office within Human Resources, transforming HR into a data-driven organization with a focus on data governance, security, management, and compliance for all HR, cyber, and legal data across global SVB. She is responsible for implementing and executing retention and purge via classification in HR while leveraging the ServiceNow platform.