Chatbots – a Double-Edged Sword

Guest post by Francis O’Haire, Group Technology Director, DataSolutions

Artificial intelligence (AI), especially when it comes to chatbots, is garnering a lot of media interest at the moment. When OpenAI (a US-based AI research laboratory) launched ChatGPT back in November 2022, it was seen by many as a significant breakthrough in technology – AI capable of providing users with detailed, coherent responses spanning a wide breadth of knowledge. But like all ground-breaking advances in technology, there are both positive and negative use cases – splitting the atom, for instance, offered both the promise of cheap energy and also the threat of nuclear devastation.

Indeed, even the leaders of great tech companies have arrived at a crossroads in their respective views concerning AI, with Bill Gates criticising Elon Musk and more than 1,000 tech leaders for wanting to halt the development of AI due to fears it could destroy humanity. Gates is of the opinion that pulling the plug on the evolution of ChatGPT and other similar systems will not solve any of the problems or challenges surrounding them. Despite which side of the fence you’re on when it comes to chatbots, the reality is that they do represent a double-edged sword.

Europol recently released a report which shows how ChatGPT could be used for criminal purposes, observing that “while all of the information ChatGPT provides is freely available on the internet, the possibility to use the model to provide specific steps by asking contextual questions means it is significantly easier for malicious actors to better understand and subsequently carry out various types of crime”.

For example, phishing campaigns often fall down or fail due to the perceived authenticity of the message with poor spelling and grammar or unrealistic branding, giving attackers away. However, technology like ChatGPT could greatly assist in addressing such shortcomings and rapidly produce much more professional-looking phishing emails which are more effective.

Also, ChatGPT can produce computer code without requiring a strong background in programming. While there are safeguards in place to limit the creation of malicious code, these are by no means foolproof. Those looking to exploit this can manipulate the questions they input to make it produce something that it shouldn’t. An example was recently documented by Check Point Research, where cybercriminals used ChatGPT to improve the efficacy of an existing piece of malware.

Then there’s the prevalence of ‘fake news’, an area where AI has proven itself very capable of generating realistic images based on the user’s requests. These can be used to promote falsehoods or spread fake news, which can be used to further the agenda of nefarious individuals, groups or even governments. Prime examples of this are the fake pictures of former president Trump supposedly being arrested and detained by the police.

And let’s not forget about deepfakes. Many of these are pornographic, with research showing that a staggering 99% of these manipulated images mapped faces from female celebrities onto the bodies of porn stars. It begs the question, will we see examples of this being used in so-called revenge porn in the future? Of course, there are also less inflammatory examples, like the fake photo of Pope Francis sporting an impressive designer coat which went viral and fooled much of the world.

The chatbot space will only become busier and more competitive going forward with Google’s launch of Bard. In turn, this will not only serve to facilitate further advances in AI technology but also assist in its use for good. In the world of business, chatbots with AI and machine learning capabilities can help organisations to redefine and enhance customer service.

They can handle thousands of customer queries instantly, as well as simultaneously improving the average response time. They can also help businesses maintain solid levels of consistency when answering queries and improve customer experience, with the capability to be programmed as multilingual, for example.

Furthermore, we are already seeing cyber security vendors using these technologies extensively to keep ahead of hackers and lighten the workload for overwhelmed cyber security teams. Advances in AI and chatbots, such as ChatGPT, can be incredibly powerful tools in helping to combat an increasingly complex and sophisticated threat landscape. In other words, given that the bad guys are finding ways to use this incredible technology, it’s crucial that the rest of us do the same to level the playing field.