ChatGPT, a machine learning (ML)-powered chatbot, is rapidly growing across all sectors.
The app's developer, OpenAI, reported that it gained one million users in just five days. The app has now been visited over two billion times, according to research by Similarweb. This being said, concerns have been raised about the use of the intelligent chatbot, with Italy's data privacy agency even going so far as to temporarily ban the use of the app in the country over concerns that it violates GDPR law.
Due to the app's impact on the sector, Cyber Security Hub’s Advisory Board members discussed ChatGPT's impact on the industry in its March meeting.
Cyber Security Hub’s Advisory Board is a group of experts in IT security and technology that meets every two months to discuss the most important issues and developments in cyber security.
How AI enhances cyber security strategy
With research by Capgemini finding that the majority (69 percent) of enterprise executives said that artificial intelligence (AI) was necessary to respond to cyber security threats three years ago, the complete integration AI within cyber security strategy seems inevitable.
When discussing whether they use AI in their cyber security strategy, however, one member was quick to point out that common misconceptions about what AI is can muddy the waters when discussing its use.
“AI is a huge buzzword at the moment but what people are talking about is not true AI as such,” they explained, “We use a lot of ML as we need to understand all user behavior analytics from ingress points through to instruction. AI is instruction-based and ML is behavior-based.”
Another member agreed that they do not use as much AI as the technology is still in its infancy, however they do use machine learning as it can use data to make predictions in a fraction of the time it would take a human to. They also noted, however, that true AI may open up more risks to a company’s network. This potential risk has led to the member “follow[ing] trends more on the conservative side, leveraging people and using technology as a blend to get the best results”.
The future of artificial intelligence and machine learning
Cyber Security Hub research has found that almost one in five (19 percent) of cyber security professionals say they are investing in cyber security controls with integrated AI and automation.
When considering how and when AI and ML will be integrated into most if not all cyber security solutions, a member said this will happen once those in the cyber security industry realize that they cannot change human behavior.
“You can positively or negatively reinforce behaviors. It is great to automate and great to use AI but it also needs the human, we should not forget that we cannot have a tool for everything,” they shared.
Another member agreed, saying that AI and ML will continue to progress in the workforce as cyber security itself has a lack of people who wish to get involved and gain experience, choosing to rely on technology instead.
“You can positively or negatively reinforce behaviors. It is great to automate and great to use AI but it also needs the human, we should not forget that we cannot have a tool for everything.”
“Innovation is becoming so critical in all areas that we need to keep pushing the needle forward. It is exciting but scary, because you can have the machine do things that usually need multiple people with the click of a mouse. What you can get from ChatGPT used to take hours or days, but people always have to be part of the process as long as there are people to do it. I don’t know if there are enough people coming in [to the cyber security space].”
Based on this, members agreed that behavioral scientists will be involved in the expansion of machine learning especially, as they will be able to drive machine learning algorithms and allow them to anticipate decision trees to quickly make decisions or provide several avenues for the decision.
With this being said, one member clarified that AI and machine learning will never truly overtake humans, even if it does manage to catch up with the speed of human thought: “AI and ML will supersede but as soon as processing power catches up to brain power it will take over. It still needs the human, however. Social media and cyber warfare will drive the AI and ML evolution forward”.
Why cyber security professionals are concerned about ChatGPT
Research by Blackberry Security has found that cyber security leaders are concerned about ChatGPT’s use by malicious actors, with 73 percent either ‘fairly’ or ‘very’ concerned about the AI tool’s potential to be used as a threat to cyber security.
When discussing this concern at the meeting, one Advisory Board member described that they had already noticed ChatGPT being used to make cyber attacks more sophisticated within their company.
They explained that they see about 37,000 phishing campaigns weekly and have recently noticed that malicious actors have gone from using broken or misspelled English to “prim and proper” language. The member suspected that they have started using ChatGPT to craft a style that helps them with their English.
The member also noted that ChatGPT is also helping malicious actors to understand the psychology of the recipient and better put them under duress to increase the effectiveness of their phishing attempts. To combat this, people have been developing anti-GPT solutions, including one that can tell whether content has been typed by a human or systematic programming.
“AI and ML will supersede but as soon as processing power catches up to brain power it will take over. It still needs the human, however. Social media and cyber warfare will drive the AI and ML evolution forward”.
Another member dubbed ChatGPT as “cool but scary” because of its potential to be used by bad actors.
“Phishing is the number one attack surface and [malicious actors] will use it to make scams more realistic. It will be the voice of spear phishing and targeted spear phishing will be enhanced due to ChatGPT. It is just another way to increase their success with their attacks.
“When you talk about malware or ransomware, bad actors [may use] third parties as ransomware, [but] now we may see them using ChatGPT and eliminating the third party. There is lots of good but there is always something bad to do with it,” they explained.
Later on in the discussion, a member noted that ChatGPT may also cause an issue within cyber security teams, as if their development team is using ChatGPT to generate code and then using this within their platform, the code may be unsafe and open their network up to a number of threats. They said that this problem may be exacerbated if companies are constantly hiring new developers, as they may feel reliant on ChatGPT to complete their work quickly.
The member explained that employees may take code from ChatGPT without reviewing it first as it is human nature to trust sources, even if these sources come via the internet. They noted that those in cyber security must move quickly to stay on top of technological changes, like the development of AI, as well as to mitigate the aspects of human behavior and psychology that are a threat to cyber security.