This article is by Tim Hardcastle, CEO, INSTANDA.
The latest Gallagher Re Global InsurTech Report includes a statistic that should capture industry attention. In Q1 2026, AI-focused Insurtech firms captured 95.2% of all Insurtech funding: Across 68 deals, they raised $1.55 billion out of a total $1.64 billion across the whole sector. Every single one of the quarter’s ten largest rounds went to an AI-focused business. Back-to-back with Q4 2025, these are the strongest consecutive quarters for Insurtech funding in the past four years.
Within that overall figure, there is a specific number that matters even more: companies operating in AI liability and cyber insurance raised $444.84 million in Q1 alone. This shows that this market is already in the process of forming at speed, attracting serious institutional capital, and building products that incumbent insurers are currently not positioned to offer.
Two markets becoming one
For years, cyber and AI liability sat as separate conversations. Cyber was operational risk, such as data breaches, ransomware, and business interruption, while AI liability was a more speculative proposition, as a possible framework in the future. The Gallagher Re report argues persuasively that this distinction is dissolving.
The risks are converging around a common problem: what happens when digital systems make decisions that cause harm, and who is accountable. An AI-powered phishing attack that triggers a ransomware event is embedded in existing cyber policy language. A hallucinating AI agent giving bad medical advice, a discriminatory recruitment algorithm, or an autonomous pricing model that systematically disadvantages a protected class, currently sit in a gap between cyber liability, professional indemnity, general liability, and product liability that existing policy language was not designed to address.
There is a further dimension that sits beneath both risk categories. Securing an AI model itself is extremely difficult, as the attack surface is probabilistic and shifting. The more practical and urgent priority is securing the perimeter around it: the data it can access, the decisions it can trigger, and the administrative frameworks that determine how it operates within an organisation. Insurers designing products in this space need to reflect that reality in how coverage is structured, not only in how it is priced.
The report is clear that AI as an attack vector typically triggers cyber coverage. AI as a liability source, through hallucinations, information bias, data disclosure, or regulatory breaches, generally does not. The gap between these two sides of coverage is where the $444 million of investment is going.
What the next generation of products must cover is not a technology risk in the traditional sense, but a statistical one. AI models produce probabilistic outputs and will therefore sometimes produce incorrect answers. This is not something that can simply be coded out, but is an inherent property of AI systems, and insuring it requires fundamentally different underwriting than covering a network intrusion or a software defect.
The convergence also has an infrastructure dimension that is less discussed but equally important. Effective cyber and AI liability underwriting will require access to a level of data granularity that today’s systems are not built to provide. To price algorithmic prejudice risk, firms will need to understand how a model was trained and on what data. To assess AI agent exposure, they will need visibility into how self-directed systems are governed and supervised. To underwrite convergence products that span identity, payment, and AI decision-making simultaneously requires the ability to connect data sources, configure coverage parameters, and iterate policy terms at a speed and flexibility that existing policy administration infrastructure cannot support.
The often-ignored operational problem
The challenge for incumbent insurers is not ambition. Most carriers understand that AI liability and cyber convergence products represent a significant commercial opportunity. The challenge is practical and structural: their existing operating models were not built for this class of risk.
AI liability is cross-functional by its nature. Pricing a probabilistic, model-dependent exposure requires underwriting, claims, actuarial, and data functions to work from real-time information. Yet most incumbents maintain siloed data pipelines across these functions, as a legacy of traditional insurance organisation. This made reasonable sense for static risk classes but becomes an operational liability when the risk itself evolves more quickly than the internal processes can track. The cost is not only speed, but is the organisational weight of reconciling data, managing handoffs, and maintaining coherence across teams who are working from different information at different times.
There is also a talent dimension that the Gallagher Re report identifies but does not fully explore. The specialist AI risk underwriting expertise required to compete in this market is rare and concentrated in a small number of firms and reinsurers who have been deliberately building it. In many incumbent organisations, skilled underwriters spend a significant proportion of their time on configuration, data retrieval, and administrative reconciliation that adds no underwriting value. The tools they work with were designed for a different era of risk and force them to absorb friction that automation is better placed to handle. The result is that rare human expertise is being consumed by tasks that do not require it.
This is where the human-in-the-loop argument becomes directly relevant to competitiveness. AI liability products are themselves going to require AI-assisted underwriting, as probabilistic, model-dependent risk at scale cannot be priced with a purely manual workflow. The skilled professional’s role will therefore shift to reviewing model outputs, interrogating edge cases, and applying contextual judgement where the algorithm cannot. That is a more demanding job, and it depends on removing the administrative burden that currently consumes the capacity needed to do it well.
Platforms that eliminate the friction connecting data sources, automating routine configuration, and surfacing the information underwriters need to make decisions, will not only accelerate product launches, but will also make the humans doing the skilled work more productive and focused, and easier to retain in a market where that kind of talent is already scarce.
The window is still open, but will not be forever
For incumbents, the window to participate is still open, but is rapidly narrowing. The Insurtech capital flooding this space is not just funding new AI offerings, but also buying time and experience in the sector. Every quarter specialist players such as Munich Re’s Insure AI team and Stoïk spend in the market, they are accumulating claims experience, refining underwriting models, and building proprietary datasets around a class of exposure that is only going to grow as enterprise AI adoption accelerates.
The question for incumbent carriers
Incumbent carriers must question whether their current infrastructure and operating model can deliver more competitive offerings before that window closes. For many, the answer requires confronting two separate problems at once: the technology, which lacks the flexibility to configure and deploy novel products quickly; and the organisation, which has not yet adapted workflows, data architecture, or talent model to the demands of AI risk.
Neither problem requires wholesale rip-and-replace of entire systems as its solution. Core systems can be repositioned as systems of record, with modern orchestration layers providing the configurability that existing infrastructure cannot. Operating models can be redesigned around cross-functional data access without dismantling existing structures, and underwriters can be redeployed toward higher-value work as automation absorbs the tasks that have historically consumed their time.
The market will not slow down while carriers plan their transformation. The practical route is to act on what is available now: improve the configurability that is missing, redesign the workflows that are creating drag, and empower the people whose judgement the industry ultimately depends on. The claims experience that sharpens underwriting only comes from participating in the market, even imperfectly. But the carriers most likely to compete effectively are those who arrive with both a product and an operating model built for the nature of the risk they are now being asked to cover.
Be the first to comment