Cyber insurance is a complex and fast-paced area of the insurance sector, which is constantly changing as cyber technology develops. Cyber innovation has always posed both risks and benefits to businesses, with insurance companies working hard to stay a step ahead, building insurance policies that respond to new innovations appropriately.
One such recent innovation is a market-leading AI model - ChatGPT.
What is Chat GPT?
Developed by OpenAI, ChatGPT is an AI chatbot that surpasses many previous chatbot types, thanks to a Generative Pre-trained Transformer (GPT). The difference from older chatbot types is that ChatGPT can analyse context and wording to create complex and relevant conversation and prose. ChatGPT has “learned” (or been “trained”) using language from huge amounts of internet-based data, in order to recognise a wide range of different ways in which texts are formed and interact.
OpenAI explains that in addition to the information-based training, it has also used Reinforced Learning from Human Feedback (RLHF), which included human trainers ranking AI responses by quality to create “reward models”.[i]
The way that ChatGPT has been trained allows it to create answers and content that is very human-like. Having a chatbot that can imitate human reactions offers numerous possibilities for businesses and individuals.
As many clients start to pay attention to what ChatGPT is offering, the insurance sector needs to consider how it could affect their risk profile.
What are the benefits?
The benefits of ChatGPT for any business is relatively easy to imagine. It can have an impact on everything; from marketing content, to customer services, professional emails, and gathering content for blog articles. The AI system can analyse huge amounts of data in microseconds, allowing almost immediate responses that can be accurate, relevant and appropriate.
In customer services, in which chatbots are already in mainstream use in some sectors, ChatGPT can emulate human responses to queries and amend its responses as the customer clarifies their questions. It can even challenge questions or comments that are not correct or that are inappropriate. If it is used for straightforward interactions, it can free up time for human customer service representatives to focus on more complex tasks.
It can support professional communications, developing well-articulated paragraphs at the touch of a button. It can decrease the time spent on research and analysis for content writing or information gathering. For marketing it can create prose that flows accurately and reads well; ideal for content creators wishing to develop articles on topics.
What are the risks and drawbacks?
As with any new cyber technology, ChatGPT does also come with a new range of risks and drawbacks.
The dark web is already offering forums on how ChatGTP can be utilised to create phishing scams and malware, with the potential for an increase in these kind of activities from individuals who previously would not have had the capacity to create them.
[ii] It is vital that as hackers develop new ways to take advantage of ChatGPT, the developers and companies are ahead of the game with their cyber security, protecting themselves from these threats.
When using ChatGPT as part of their business strategies, companies need to seriously consider data hygiene. Managers need to think about what information they are going to input into ChatGPT. The release of classified data or trade secrets could be devastating to a company – not to mention the legal repercussions of such a breach. There will need to be policies in place for the decisions around what content will be fed into the AI.
The information ChatGPT can offer is only as good as the information available. Most people would acknowledge that the internet is not wholly a source of correct, full information. Therefore, whatever information is gathered needs to still be analysed by someone who is an expert in the field. It is also (currently) limited to information up to and including the year 2021 as the RLHF training completed in early 2022. It is not, therefore “up to date” on current world events, which can be limiting.
As a “learning” AI, there is the risk of bias. If the system has had access to large amounts of text with one particular bias – for example, a political bias – it is possible that the responses it generates to some questions may hold that same bias, leading users to understand that the company also holds that position. Again, policies and good management need to be utilised to mitigate these risks, as well as ensuring human oversight.
With the risks from ChatGPT, as with the development of any new cyber technology, insurers need to build cyber insurance policies that are fit-for-purpose, while also demanding realistic levels of risk management from the companies in question. As specialist insurance recruitment agents, Aston Charles are working with a wide range of candidates with experience at the cutting-edge of disruptive technologies. These individuals are able to support insurance businesses as they seek to work with clients to mitigate risks and create insurance policies that meet their needs. It is vital that insurers continue to offer products that are versatile, relevant and constantly changing to keep up with new technology as it arises.
Can insurers use ChatGPT?
As well as looking at ways in which ChatGPT offers risks and benefits to its clients, insurance companies and brokers should be asking how they may benefit directly from ChatGPT. For the insurance sector there are a wide range of ways that it could be beneficial.
ChatGPT could be used in marketing, customer services, claims and risk analysis – if implemented in the right ways. Make no mistake; AI cannot take over from human interactions, as qualified, experienced staff members are needed to assess the value of the output being given by ChatGPT. The real benefit of ChatGPT may be in giving time back to Underwriters, Risk Analysts and Customer Service representatives to focus on tasks that really require their expertise and skill.
Is ChatGPT the future?
As with most new technologies, ChatGPT can certainly be utilised for the benefit of insurance businesses and brokers – as well as for their clients. However, there needs to be competent oversight in how it is used, and it is vital that there is adequate protective technology in place as ChatGPT and similar technology continues to advance.
[i] OpenAI. 2022. Introducing ChatGPT. [ONLINE] Available at: https://openai.com/blog/chatgpt. [Accessed 2 May 2023].
[ii] Insurance Business, David Saric. 2023. Four things insurers and brokers must know about ChatGPT. [ONLINE] Available at: https://www.insurancebusinessmag.com/us/news/breaking-news/four-things-insurers-and-brokers-must-know-about-chatgpt-441555.aspx. [Accessed 2 May 2023].