AI boom could lead to new ‘silent cyber’ crisis for market

Underwriters are concerned that the increasing use of artificial intelligence (AI) systems by businesses in the development of their products and services is significantly increasing their risk of intellectual property (IP) rights and patent infringement. They also warn that the widespread use of AI is creating an uncomfortable dynamic in the market, similar to the silent cyber issue, which has led to disputes between businesses and their insurers in recent years.

This dynamic already exists in the market, according to Madeleine Brown, head of intellectual property risk at CFC Underwriting. A lot of commercial insurances contain IP infringement coverage, she says. “But the extent of IP coverage contained in these policies varies dramatically. As insurers, we need to understand the exposures and risks AI brings and the intent of the policy. At the same time, every business should consider whether their insurances are affected by ‘silent IP’”.

The risk exposure has also increased because the application of AI has become more front-facing in the sense that businesses are using the technology to drive customer chat systems, to invent or develop their own products, or to solve complex coding problems.

This has resulted in an increase in AI-related IP claim notifications, Brown says. “In regards to patents, it is currently more of an arms race than anything else but, as with any fast-developing sector, there will be claims as businesses seek to exploit and to protect their IP rights. It would therefore make sense for any businesses exposed to AI to explore purchasing a standalone IP insurance policy, which covers all types of IP disputes,” she adds.

A specialist underwriter focused on emerging technologies, CFC has been writing elements of cover for AI exposures over the past decade or so. It is a risk the company is constantly reviewing. The key difference now for CFC is that AI is at the top of the agenda for many businesses and industry sectors, driving important developments in product and service innovations, particularly in the tech sector.

Key risks

The threat landscape is varied, but for CFC, there are two key risks attached to AI. One is in relation to generative AI where the technology as able to produce texts, images or other products from prompts by a human user. These prompts could potentially be unauthorised use of copyrighted materials to train AI to generate output. “This in itself could then expose works protected by IP rights and, in turn, become an infringement issue,” Brown says.

'The extent of IP coverage contained in these policies varies dramatically. As insurers, we need to understand the exposures and risks AI brings and the intent of the policy'

^{Madeleine Brown, CFC Underwriting}

The other big risk for businesses is using AI to develop their existing products. If AI is not properly trained or programmed then it may inadvertently infringe on third-party IP rights. According to Brown, this needs to be treated like any other case where a business is looking to extend the scope of existing commercial applications, products or services. “The proper due diligence has to be undertaken to ensure you have the right to operate. There is a risk that where a human has not developed and assessed fully the training of the AI system, the technology may have utilised third-party IP without consent,” she says.

Then there is the question of ownership issues arising from collaborative agreements, including disputes over who owns what the AI has developed. This makes contract breach a very real and potentially damaging risk.

For Brown, these issues raise fundamental questions in relation to AI and copyright. For example, are AI-produced works truly original and how will these be protected? At the moment, there is no one answer, not least because IP law varies in every territory. In the US, for example, only inventions created by “natural persons” can be awarded patents. In most other territories, however, including the UK, this is not the case. In these territories, an industrial or other process generated by AI is not excluded and is patent-eligible. “It’s crucial to be aware of your exposure if you are selling your product/solution to a third party in another country and what protection you have in that country,” Brown says.

The main point for Brown, however, is that while the IP insurance market may not have all the answers given the rapidly developing nature of the risk landscape, insurers must start talking to clients about how they are using AI, and to what purpose and benefit. They then need to discuss with clients how they can protect themselves against this risk now.

At present, the IP market is dominated by large corporates and the costs of defending or pursuing any IP litigation are high. Brown does not see this changing when it comes to IP relating to AI. “This could leave small to medium-sized enterprises (SMEs) vulnerable to infringement, which is where an IP policy is invaluable. For an SME – they really need to be able to defend and pursue to fully exploit their rights,” Brown says.