Tech leaders’ perspectives on implementing ChatGPT and the looming cybersecurity risks
The general public’s curiosity about what might be achieved with AI has been stoked by ChatGPT’s sheer power and capabilities. With more than a million active users and growing, ChatGPT has captivated the world with its ability to be used in a wide range of contexts, including composing essays and articles, brainstorming based on original questions, and developing code in popular languages like Python and SQL.
A growing number of companies are starting to use it internally for tasks like information organization and research, business plan drawing, sales copy optimization, and more.
What do leaders think about it?
Tech leaders are hopeful that ChatGPT would offer fresh perspectives and support value development. Yet, there is still a substantial amount of work to be completed in terms of implementation and validation; success will be fueled by deep domain expertise, distinct use cases, and value models.
“We use several applications of AI to provide ease of use for our products, and we see AI being useful for businesses across industries going ahead. Large Language Models (LLM) are good for both, B2C and B2B, to enable conversational interaction in our products,” said Jaiteerth Patwari, Director of Engineering, Uber Delivery.
We human beings want to interact with humans like conversation. ChatGPT allows exactly that, making it a real human bot. In its current form, it may not recognize people with good or bad intentions. As things evolve there will be more guardrails of what can be done and what cannot be done.Jaiteerth Patwari, Director of Engineering, Uber Delivery
ChatGPT represents a significant improvement in the language skills of AI models. It is an LLM that can comprehend and produce text in natural language.
“ChatGPT is a leap forward in AI-augmented search. The conversational answers can be built upon with more specific inputs, will ease information aggregation, and will be a great productivity enhancer across the entire gamut of the information world,” mentions Gaurav Chaudhri, CTO, Reliance General Insurance.
The ChatGPT engine can be made to learn and build cross-functional knowledge across underwriting, actuarial, finance, servicing, and of course technology teams. Which in turn will help take improve decisions making and later down the line help automate a lot of decisions.Gaurav Chaudhri, Chief Technology Officer, Reliance General Insurance
In the upcoming months, ChatGPT’s function in conversational AI will be to assist and close the communication gap between humans and machines, enabling more purely organic interactions between people and AI systems.
“Although chatbots have been in existence for a while as a critical component of new-age customer experience, with cutting-edge technologies such as ChatGPT, they are going to get more mainstream and witness rapid adoption,” Gaurav Kachhawa, Chief Product Officer, Gupshup said.
In the future, with ChatGPT’s advanced technology, we will see a proliferation of use cases as more and more companies look to offer enhanced customer experience. We are already seeing a lot of willingness among companies to include Auto Bot Builder as a part of their customer experience stack.Gaurav Kachhawa, Chief Product Officer, Gupshup
“It remains to be seen how these engines can be made available for captive use. While NLP (Natural Language Processing) based needs span from servicing customers, let us not forget how it can greatly help in fulfilling the training needs of employees and customers,” said Chaudhri.
ChatGPT deployment will soon surge, changing how organizations interact with AI, from accessible applications like creating high-quality chatbots to more creative ventures.
“Another thing that differentiates ChatGPT is its ability to log context from users’ earlier messages in a thread and use it to form responses later in the conversation. So, this is very different from how computers used to behave before,” said Jijy Oommen, CTO, Aavas Financiers.
With its advanced language generation capabilities, ChatGPT can be integrated into a variety of applications and digital assets.Jijy Oommen, CTO, Aavas Financiers.
As more businesses strive to provide improved customer experiences, we will witness a profusion of use cases in the future thanks to ChatGPT’s advanced technology, the tech leaders feel.
Will it aid the cybersecurity industry?
The past teaches us that even when cutting-edge technology is created with the noblest of intentions, it will inevitably be put to bad use. AI is by no means an exception. Yet, both the beneficial and harmful applications of the technology seem to have been stepped up with OpenAI’s ChatGPT. There is growing evidence that the AI-powered chatbot might be a potent tool for cybercriminals and cyber defenders when it comes to cybersecurity.
“Natural language processing (NLP) models can be used for identifying and classifying cybersecurity threats based on language analysis of sources such as phishing emails, social engineering attacks, social media, online forums, news articles, etc,” explains Nick Parrotta, President, DTS & CDIO, Harman. This can help the cybersecurity industry to build early warning systems to identify system-related threats and vulnerabilities.
Algorithmic and data bias are key risks to be considered while using large language models, as the output language generated by the model could be biased, inappropriate, or inaccurate and is a function of the input training data. Enterprises need to ensure that the model is trained on diverse and representative data and protects sensitive data.Nick Parrotta, President, DTS & CDIO, Harman
“ChatGPT can also be used by defenders to create detection like Yara or Snort rules (YARA is made primarily for scanning files and/or memory, Snort is made specifically to scan network traffic) or use it to detect anomalies inside logs,” said Candid Wuest, VP, Cyber Protection Research, Acronis, and Security Advisor to the Swiss Government.
AI models can also be employed to determine whether an AI program wrote a phishing text. In conclusion, ChatGPT can assist defenses in automating and streamlining their detection chain.
ChatGPT will enable more people to “do” cybercrime, at least on the lower/basic level, which can still be successful. This will generate more attack volume, but not necessarily new attack methods or highly sophisticated attacks, as AI does mainly combine existing work and not invent new stuff.Candid Wuest, VP of Cyber Protection Research, Acronis and Security Advisor to Swiss Government
Potential cyber risks
As chatbots and AI-powered technologies proliferate, cybersecurity will face new difficulties, particularly if these tools fall into the wrong hands. Several concerns about possible risks are emerging since ChatGPT can generate programming codes at will.
We are only beginning to scratch the surface of the possible benefits of adopting this kind of modern technology, but we also cannot disregard the drawbacks. It will get harder to defend without deploying AI to equalize the playing field as the platform develops and hackers gain experience.
“ChatGPT makes AI more accessible, hence more people with no technical skills can join cybercrime. Although the code quality generated by auto-pilot and ChatGPT are not great, various tests have shown that they make mistakes as well, so there will be a need to fix them (or train AI better),” Wuest added.
Here are some of the potential cyber risks associated with ChatGPT:
- Composing plausible phishing emails, particularly for BEC (Business email compromise), even though it presently lacks an Internet connection, so if attackers want to modify emails, they must be integrated with LinkedIn, etc.
- More non-English speakers are equally at risk due to improved regional coverage of local languages through ChatGPT
- Source code analysis to look for flaws, such as buffer overflows and SQL injections.
- Generate malware that is only partially banned as per the cyber rules, but basic macros are still feasible. Anyone can find them on Google. Thus, making it more accessible to the general public.
“ChatGPT is currently offline and its learning is back-dated, still, the battle for protecting the industry from cyber-attacks will need to pick up pace as this matures. This will allow novices with limited hacking skills to try and experiment. Coupled with the advent of Quantum computers, security firms, and the industry will find it an extremely uphill battle to keep pace with the threats,” observes Chaudhri.
Note: This article is 3rd in line of an ETCIO series on ChatGPT