James Watt innovated on the Newcombe steam engine by placing a separate condenser for which he was granted a patent in 1769. This innovation increased the efficiency of the steam engine by 75 percent.
It took another seven years for first Watt steam engines to be commercially produced and installed. Little did he comprehend that his mechanical muscle would trigger the Industrial Revolution.
In today’s world, Generative Pretrained Transformers or GenAI is expected to unleash much wider creative destruction, causing a tectonic shift akin to what experts’ terms as 4th Industrial Revolution (4IR).
A 2025 Delloite study indicates that successful digital transformation can result in up to $1.25 trillion in additional market cap, and GenAI is proving to be a powerful accelerant.
Various studies conducted by big consulting companies like KPMG, PWC, Mckinsey, Gartner reveal a constrained optimism regarding adaption of AI by CEOs around the world.
These studies reveal paradoxes between intention and operational maturity, while globally 100 percent of CEOs surveyed intent to invest in AI in some form or another, but at the same time majority acknowledge uncertainties on the profitability of these investments.
If one considers AI as an arena of life, it’s a classic situation as described in ancient Indian text, Geeta, "action is better than inaction".
Although CEOs are more optimistic about the AI investment pay offs, almost all of them have concerns about cyber security and privacy risks (intellectual property (IP) infringement), lack of understanding of how AI makes decision (potential for hallucinations, bias among others) and its implications and technological failure.
In this background, as enterprises move to incorporate AI in the core of their operations, there is a "trust gap". Now, if there is a trust gap, what is fueling adaption of a still emerging technology. What is the business model in play by AI vendors?
I first heard about bundling of indemnity with AI products as a new business model from Prof Karan Girotra of Cornell Tech. My further research reveals that vendors have developed a toolbox consisting of indemnity arrangements, performance guarantees and specialised AI insurance products backed by global insurance players to transform high risk AI products into bankable assets.
One of the significant legal headaches in Gen AI applications is IP Infringement for both users and developers arising out of training data sets on copyrighted materials without authorisation or if the output may be substantially similar to protected works, which may cause direct infringement, or "contributory or vicarious infringement" if they facilitate or benefit from such activities.
Also, there are issues on human authorship and copyright protection on the AI-assisted work outputs without meaningful human input. These further compounded by plethora of global privacy laws is legal minefield.
Several global IT giants as well startup companies have prepared a counter for this with programmes like Google's "Risk Protection Programme", Adobe Firefly's "Commercial Safety", Microsoft’s "Copilot Copyright Commitment" by partnering with global specialty reinsurers.
These work by undertaking to defend the client and paying any legal settlements, if sued for copyright infringement based on the output of an AI tool.
As reported earlier on the apprehension of the profitability and outcome of AI investments, many AI companies have come up with outcome-based guarantees which takes it one step beyond performance guarantees in a bid position GenAI as "Guaranteed AI".
In these cases, if the vendor is unable to deliver key performance indicators (KPIs) which it had promised and contractually bound, it pays a penalty to the client, backed by a third-party insurer.
The active indemnification response by the insurance industry has now become a backbone for economic confidence which is an active ingredient for scalability and adaption of AI, thus paving way for its industrialisation.
The insurance industry’s response to the emerging risks by AI has resulted to a new class of specialised insurance products with "Affirmative AI Coverage".
The traditional insurance policies fall in the "Silent AI" category, which either specifically excludes liabilities arising out Gen AI applications or, the client (insured) hopes that the broad definitions in the insurance policy will cover.
This has left AI companies and their clients completely exposed or with severe coverage gaps. Take a case of errors due to hallucination by an AI model. The insurance company can argue that a traditional professional indemnity will cover errors by human but not algorithmic errors.
An insurance policy with "Affirmative AI Coverage" with specifically cover AI loss scenarios like algorithmic error and omission, model drift and performance discrimination and bias, and IP Infringement among others.
AI companies are actively bundling these insurance products in their offerings for sales acceleration and regulatory alignment. The startup AI companies are relying on insurance for their own reliability to scale their offerings and compete against Big IT giants for enterprise contracts.
These has also provided the confidence to clients to deploy high stakes AI projects in their respective organisations, knowing there will be financial indemnification if the technology fails.
The writer is a re-insurance executive focused on AI-led transformation and Chief AI Officer from Cornell University.
Unlock a world of exclusive content today!Unlock a world of exclusive content today!
