Log in

Forgot password?

Don't have an account with us?

Register Now

Generative AI in Practice – Issue 4

Generative AI in Practice is InsTech’s monthly newsletter dedicated to the use of generative AI in insurance.

The InsTech Perspective

OpenAI’s view of catastrophic risks

Last week OpenAI, the company behind ChatGPT, announced the launch of a ‘preparedness’ team. The division will ‘track, evaluate, forecast and protect against catastrophic risks’ from advanced AI models such as its own GPT-4. OpenAI lists four examples of such risks:

  • ‘individualised persuasion’ (generative AI creating convincing personalised messages that manipulate people);
  • cybersecurity risks;
  • chemical, biological, radiological and nuclear (CBRN) threats; and
  • ‘autonomous replication and adaptation’ (AI going rogue and humans losing control).

We believe insurers should follow OpenAI’s preparedness research closely. In particular, if your company is one of those that insure OpenAI, Google or their competitors, the liability risks could be very large.

Even if you don’t insure AI companies you are still at risk. Your policyholders could be exposed to AI-enabled manipulation, cyber attacks or CBRN events, which may be covered under existing insurance policies. On the other hand there will be opportunities to diversify by creating new insurance products for AI risks (see below).

OpenAI is also asking the public to help it understand risks; it has launched a competition for people to submit their views on AI threats and mitigations. We hope people across the insurance industry will participate. Our industry, with its experience modelling natural catastrophes, cyber and other risks, should be leading this conversation.

Use case of the month

The opportunity to insure AI

Companies across many industries are increasingly using generative AI for operational efficiency and to build new products and services. These expose organisations to new risks, some of which are insurable.

Since 2018, Munich Re has offered aiSure, as insurance protection for AI companies that offer performance guarantees to their customers. Munich Re is now extending this protection to organisations that use their own AI models in house. Last month, start-up Armilla also launched a performance guarantee for AI products backed by reinsurers Swiss ReChaucer and Greenlight Re.

Meanwhile, GoogleMicrosoftAdobe and IBM are providing some legal protection to their generative AI customers. For example, Google will assume responsibility for the legal risks involved if you generate something using Google’s AI models and are challenged for copyright infringement. It’s not clear if these guarantees are backed by a traditional insurance structure, but they could be. With big tech companies fiercely competing on their generative AI solutions, performance guarantees are an important way they can gain customers’ trust.

An opportunity for insurers right now is to underwrite guarantees that are wider in scope than those already on offer. Providing such protection is likely to continue to be a competitive advantage for AI companies for some time.

What the insurance industry is doing with GenAI

Claims teams at Travelers are piloting a GenAI knowledge assistant to help access claim information quickly when speaking to customers.

Liberty Mutual has worked with Microsoft to set up its own instance of OpenAI’s GPT. The company is experimenting with employee productivity use cases, while introducing generative AI training for its top 2,400 employees.

Insurer Nationwide is using its claims data and generative AI to provide personalised information about pet health risks to its pet insurance customers. It is also experimenting with AI copilots for its customer service and sales teams.

Insurer Allstate has developed a GenAI tool to help policyholders report claims more quickly. The tool summarises the details of the incident and passes it onto adjusters so claimants don’t have to answer the same question twice.

Reinsurance broker Inver Re says generative AI is helping it understand its clients’ risks better, including identifying hidden inflationary effects on one insurer’s portfolio.

Canadian insurer Sun Life is using AI to write code, generate marketing content and images and detect fraud.

MGA Ridge Canada is using GenAI to reduce the time spent analysing legal documents.

The Savings Bank Mutual Life Insurance Company of Massachusetts has licensed ChatGPT Enterprise from OpenAI so its employees have access to the chatbot.

Swiss insurer Die Mobiliar has introduced an in-house ChatGPT for its employees in partnership with Microsoft.

News from the InsTech network

Capgemini is investing $2 billion in generative AI over the next three years. Insurance Leader Kristofer le Sage de Fontenay joins the InsTech podcast to discuss how GenAI is solving pain points for its customers.

EY surveyed executives at 24 European insurers. 83% expect a significant impact to their workforce and operations and 54% plan to integrate AI training into their graduate programme.

An underwriter at Liberty Specialty Markets argues that AI is an underwriter’s friend, not foe.

Sapiens has announced a new AI tool to help insurers create decision models, using technology from OpenAI and Microsoft.

Simplifai writes about eight opportunities AI presents for insurers, with a case study of how it saved a Dutch insurer 80 hours of manual processing time per week.

AI won’t replace underwriters, but underwriters using AI will replace underwriters who don’t use AI.’ – Indico Data’s CEO, in conversation with Microsoft.

Many of Cytora’s clients are using large language models in production. Its CEO explains how generative AI is changing underwriting and how to ensure data is kept secure.

Mastercard has expanded its consulting services to help companies adopt relevant and responsible AI strategies.

Markel’s digital magazine The Cover asks whether Britain will be the home of AI regulation.

Are customers comfortable with insurers using AI? Sprout.ai’s survey looks at the factors that make people sceptical and how to address their concerns.

Other news and insights we’ve seen

70% of Americans would not be comfortable with an AI insurance agent, according to a survey from agency GetSure. They were mostly worried about inaccurate recommendations and AI taking jobs from humans.

Can big tech’s AI predict storm landfalls better than conventional catastrophe models? After one AI model predicted Hurricane Lee’s landfall surprisingly well, our CEO Matthew Grant opened up what turned out to be a lively debate responding to Paul Carroll about this topic – join the conversation on LinkedIn.

AXA’s annual future risks report has featured AI for the first time, as the world’s fourth biggest concern (as ranked by experts), after climate change, cyber risks and geopolitical instability.

How InsTech has used AI this month

InsTech’s events team is using Claude, a ChatGPT alternative, to help draft engaging titles and descriptions for our upcoming events.

Working with our sponsor, our team develops an outline of the event’s theme and content. They enter these notes into Claude and ask it to generate a list of possible titles for the event that will catch people’s attention.

We also use Claude to create short, easy to digest summaries of what the event will be about.

With the help of Claude, we’ve been able to publish the registration pages for our events quickly and make it easy for people to access the information they need about what’s coming up. Evening event registrations are now getting close to 300 people per event.

Find out what you’ve missed…

Issue 3: Claims automation, billion-dollar spends and AI-generated business plans

Issue 2: Distribution at scale and GenAI use cases from Lloyd’s syndicate MIC Global

Issue 1: Extracting data from emails and Generali’s three factors to consider with GenAI

Generative AI in Practice is InsTech’s monthly newsletter dedicated to the use of generative AI in insurance – you can sign up for free here.

Stay up to date with the latest from InsTech