Open Menu
×Close Menu

ARTICLE 

AI Chatbots & Services: The New Insider Threat

HubSpot Assets Read Time Graphic (7)

Article Written: 04/01/2023
 
Disclaimer: This article was written 100% by a human.
 
We have seen a new tech craze in the last few months with the rise of Artificial Intelligence (AI), most notably ChatGPT. ChatGPT can be charming, funny, insightful, and fun to play with. It is almost everything you wanted out of your phone assistant, search engine, or the home electronic personal assistant that sits on your counter. We will likely see home and phone assistant programs even further improve their AI Natural Language Processing (NLP) technology in the next few years to the point where it will feel like talking to an intelligent person more than a clunky web search engine. Microsoft Bing has already jumped ahead of the game by introducing a Large Language model you can opt-in to use.
 
However, privacy advocates are concerned with the mass amount of data chatbots and other AI systems are collecting, and questioning how it will be used other than for Machine Learning (ML). 
 

Another example of AI-based machine learning is the recommendation systems used by e-commerce websites such as Amazon and Netflix to suggest products or movies based on your previous browsing or viewing history.

Netflix LogoAmazon Icon

Article Sections

Article Sections

How Does ChatGPT Work?

With the rise of AI Large Language Models comes a useful and powerful tool that companies can harness for productivity--but at a potentially dangerous cost. Microsoft and Amazon recently reaffirmed to their employees the importance of caution around sharing sensitive information with ChatGPT in case it’s used for future AI training models.
 
ChatGPT is very good at being an assistant for programmers. Almost like a Sous or Jr. Chef that provides the Head Chef with remedial tasks. After all, while ChatGPT can do some code, it can't do much beyond giving you a base to modify and tweak code it generates or code you have done yourself. This will soon be a problem as coders ask ChatGPT or similar programs to clean up their proprietary code. Not only can this be stored in a backend database for further research per Open AI’s Terms of Agreement, but the learning model can learn proprietary techniques and methods from programmers by changing the feedback system which is the core of its machine learning engine.
 
At the time of writing, ChatGPT claims it does not store inputted information. So, how does it learn to give better answers and ask questions? To simplify the answer, ChatGPT was started with a base-level database consisting of an unknown amount of data gathered from the internet and other sources. The internet itself is estimated to be around 5,000 Petabytes! That’s about 5,242,880,000 Gigabyte SSD Drives if you are thinking of upgrading. It was then pre-trained with a probability and weight system. When you ask it questions, it will answer based on the information it has with the highest weight score. But sometimes we have to adjust our questions or tell it that it’s wrong. It will then start gradually adjusting its scoring system to provide better answers. But the base data never changes. It can’t tell you or formulate any data about any events after September 2021.

The American Bar Association reveals 25% of firms have experienced a data breach (2)

Potential Security Risks with AI Technologies

Concern over data privacy online is nothing new. Nor is it unfounded. For years, we have seen applications from search engines, email services, and social media sites harvest personal and corporate information. And, we have seen data leaks from these services being breached and offered for sale on the dark web.
 
Do I believe there is an actual threat today? Well, yes and no. While we have no insight on what OpenAI, the company that makes the OpenAI GPT-3 engine which ChatGPT runs on, is doing versus saying, we all did agree to the license that says they can store our data for training future versions.
 
Will we become complacent with the way it might be entering corporate secrets, code, and application passwords? How will we know if OpenAI does start collecting the data? What happens if they experience a breach?
 
This is the threat. The future. It’s enough to have made leading tech experts such as Elon Musk (CEO of SpaceX, Tesla, and Twitter) and Steve Wozniak (Co-founder of Apple) nervous enough that they and others have signed an open letter to pause AI development. Their letter sights some strong language including, but not limited to, “AI systems with human-competitive intelligence can pose profound risks to society and humanity…” and “… advanced AI could represent a profound change in the history of life on Earth, and should be planned for and managed with commensurate care and resources.” and “we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors.”
However, while the hype and focus are on ChatGPT, there is another risk which is that their GPT-3 and GPT-4 engines are already being used by other companies to chat and advise users as well as other Large Language Model engines. These companies and engines have their own terms of agreement and can collect and store your data.
 
The overall risk is that some companies have disclaimers in the fine print that all data belongs to them—an existing concern many corporations have about using online presentation and storage tools.

Legal Icon

AI Lawyer: A ChatBot GPT engineer to help you with advice on legal issues.

Risk: Company secrets or personal information can be stored in a database and possibly subpoenaed or seized by federal agencies or even used as extortion by threat actors if there are data breaches.

Collaboration Assessment (9)

Ludo AI: An AI platform that helps developers with Game Creation such as mobile and casino games.

Risk: Intellectual property can be stored in databases that are capable of being breached, as well as Terms of Services giving companies rights over your IP.

Collaboration Assessment (8)

Nano Nets: Automates manual data entry with AI.

Risk: Your documents can be stored in databases that are capable of being breached as well as their Terms of Services giving them rights over your intellectual property.

Embracing Possibility, Mitigating Risk

While AI assistant models will be very helpful assistants in performing our jobs and seeking data, a danger that has always existed with other online services still looms. With AI assistant models, we will see more and more company intellectual property being transmitted to third-party services because, to your employees, it makes work easier.
 
So, we must start educating our users now, develop policies, and monitor usage. New laws, compliance standards, and best practices will be right around the corner. Do not let AI get ahead of your policies. Start now!

Event Page Icons (600 × 600 px) (24)

HS CTA Gaming eBook (1)

A Guide For How Gaming Organizations Can Tackle I.T. and Security Hurdles

Today’s digital landscape has changed how the gaming industry operates, introducing new platforms, devices, and even business models. With the rise of online gaming and the exponential popularity of smartphones, gaming organizations need to adapt, pivoting with consumer behaviors. But this technological shift has also opened the door to more cybercrime than ever.


Many key players in the industry aren’t fully aware of how cyber threats have grown, both in severity and frequency. They may have IT staff in place to respond to threats, but no way of knowing if their defenses are really able to handle new challenges. 


Learn how to strengthen your defenses against modern cybercriminals with this eBook.

 

GET YOUR COPY

Share This

Call Us

1.866.328.5538

© 2025 BULLETPROOF, a GLI Company Headquarters