Thoughts on The Use of ChatGPT in The Security Industry
Jon Polly, PSP, SICC, IC3PM
Security Technologist, Smart City Security Strategist
Not long ago, the world was taken by storm by OpenAI's ChatGPT platform. Almost instantly, laborious tasks were reduced to seconds. To say the least, the ChatGPT product was an overnight success; and an immediate concern. Following the release of ChatGPT, schools from K-12 to Higher Education immediately banned the use of ChatGPT in fear of students using the ground breaking AI as an easy button to write papers.
That action brought to light that ChatGPT is a very powerful tool but without oversight and restraint, it can easily be abused.
ChatGPT is a powerful tool, but it is only a tool.
The Security Industry has been polarized by responses to ChatGPT specifically. While many of these are touting the amazing capabilities of ChatGPT, there are some concerns that should be noted.
Report or Procedure Writing - ChatGPT has been proven to be able to write articles, policies, and even detailed post orders. Unlike the human brain which pulls from knowledge, everything it pulls is from open-source information. There is no original content, so if using ChatGPT for article writing, you may be running into issues of plagiarism; and ChatGPT does not provide sources for where it found its information. Long term, as people will use ChatGPT and other AI sources to write; if it is a deliverable to a client, don't be surprised if they use a fact-checking tool to see if an AI wrote it. Is that a conversation you need to have? If ChatGPT is used, let it be step one, with step two being a human editor to review and rewrite the text to fit the specific use case.
领英推荐
The adoption of AI-enabled bots will give rise to software to police the use of these bots.
Video Augmentation - ChatGPT has a feature for augmenting video files. Most Video Management Systems (VMS) platforms have a watermarked solution that will flag if any AI is used to alter the video. Not really worried about that. What we should be worried about is the augmentation of video that is released into the public. Body Worn Camera (BWC) video or cell phone video that is augmented with similar technologies as Deep Fake. The court of public opinion can taint the court of law. Safeguarding against ChatGPT will be required.
Code Writing - ChatGPT can, just like the report, write code. As I understand it, it can write in most of the common coding languages. Again, it is pulling from open source codes. Here's an example of the dangers. Most of the technology companies; including iCloud, Fortinet, IBM, Microsoft, Amazon, Cisco, Red Hat, Nutanix, VMware, and many more used Log4j. Log4j is an open-source code to log error messages. It was such a small and simple code, why not use it? But it was open source. In December 2021, a vulnerability was found that affected more than 75% of companies globally were impacted in some way by this vulnerability (https://www.continuitycentral.com/index.php/news/technology/7113-20-percent-of-organizations-have-had-significant-impacts-relating-to-log4j). Now, if ChatGPT is coding using open source codes, then the number of vulnerabilities will rise significantly. The problem is that companies selling coded solutions may not disclose that their code was written by an AI.
ChatGPT and other AI-enabled bot platforms should be used only as a tool, and not as a replacement for original thought.
Security Operations Manager | Solving Problems, Securing Success | Skilled in Team Building, Boosting Morale, & Improving Retention | Driven by Integrity & Accountability
1 年Although ChatGPT helps me by correcting grammar and structure in my resume and other documents, it's important to note that I still need to provide the actual content or substance of the text. In other words, I must provide the ideas and information that will make up the document. Additionally, I need to proofread the final results to ensure accuracy and clarity.
Executive ? Business Strategist ? Operations & Sales ? Board Certified
1 年Well said! It is only as good as the information generated. One must have enough expertise to review said info for accuracy and proper usage.
Senior Telecommunications Engineer
1 年Joshua A. Charfauros posted about ChatGPT last week.
Senior Telecommunications Engineer
1 年I just asked it the results of my cities last mayoral election, and CHATGPT LIED TO ME. It said the current mayor won by a specific amount over someone, and I backchecked the results and the mayor ran unopposed. ??
Business Executive
1 年Great Post Jon. Try 'Crisis management plan for a college campus" creates an accurate, high-level but useless outline of what goes into a plan but you won't have an operational plan. Typing "hit home runs like Aaron Judge" gets the same response - accurate/useless instructions. The requirement to watermark AI-created content would benefit both the vendors and the users from risk.