Insure Against "AI Causing Trouble"? Insurance Companies "Dare Not Take On"

As companies rush to embrace the artificial intelligence wave, risks are quietly accumulating. Yet, the insurance industry—which ought to play the role of a risk "stabilizer"—is showing caution and withdrawal in response.

According to the Financial Times on November 24, faced with potential billions of dollars in claims risk brought by AI technology, major global insurance companies are taking action to try to exclude AI-related risks from standard corporate policies.

Industry giants such as American International Group (AIG), Great American, and WR Berkley have already submitted applications to US regulators, seeking approval to add exemption clauses to their policies that explicitly exclude liabilities arising from companies deploying chatbots, AI agents, and other such tools.

This move marks a significant shift in the insurance industry's attitude toward AI risks. Insurers are generally concerned that the decision-making process of AI models is opaque, like a "black box," making liability difficult to define if errors occur. What’s even more alarming to them is that flaws in a single AI model could lead to thousands of related claims, creating a "systemic, aggregate risk" that the industry cannot bear.

This shift comes as the costly consequences triggered by AI "hallucinations" (where models generate false information) or mistakes are no longer just theoretical. From Air Canada's chatbot inventing a discount leading to compensation, to Google's AI search function being sued for $110 million for providing incorrect information, real-life cases continue to sound the alarm.

AI Risk as a “Black Box,” Systemic Risks Cause Concern

The core reason for insurers' withdrawal is the unpredictability and potentially huge scale of AI risks.

Dennis Bertram, head of network insurance for Europe at Mosaic, a Lloyd’s of London specialist insurer, stated bluntly that AI models “are too much like a black box,” making them difficult to insure. The company does provide coverage for some AI-enhanced software, but has clearly refused to cover large language models like ChatGPT.

Kevin Kalinich, head of cyber risk at EY, pointed out that the insurance industry might be able to bear a single company’s AI pricing or diagnostic error that leads to $400 million or $500 million in losses, but “what they can’t bear is one AI vendor’s mistake causing 1,000 or 10,000 claims—this is a kind of systemic, correlated aggregate risk.”

Zurich Insurance CIO Ericson Chan also believes that AI risks involve multiple parties such as developers, model builders, and end-users, with a complex chain of responsibility and a potential market impact that “could be exponential.”

The actions of insurance giants are particularly direct. According to filings submitted to regulators, WR Berkley's proposed exemption clause is extremely broad, aiming to exclude claims stemming from “any actual or alleged use” of AI technology. AIG also told regulators in its filings that generative AI is a “broad-ranging technology” whose potential to cause claims “may increase over time.”

These moves come as AI risk events happen frequently. Apart from previous cases, UK engineering group Arup lost $25 million last year after fraudsters used digital clones of executives to order a transfer during a video conference. As insurers seek exemptions, risks that might previously have been covered by “technology errors and omissions” policies are now facing a protection vacuum.

Seeking Workarounds, but Coverage Remains Limited

Faced with market demand, some insurers are exploring compromise solutions, but the coverage offered is often narrow and tightly restricted.

For example, insurance company QBE has launched an “endorsement” (policy amendment) for fines relating to the EU Artificial Intelligence Act, but according to a major broker, the clause caps payouts for AI-related fines at 2.5% of the total insured amount.

Zurich-based Chubb has agreed, during negotiations with brokers, to insure certain AI risks but has clearly excluded “wide-reaching” AI events—meaning scenarios where a model issue affects multiple clients simultaneously.

Aaron Le Marquer, head of the insurance disputes team at law firm Stewarts, warned that when AI-driven losses surge, insurers may begin challenging claims in court, and he expects that “it may take a major systemic event for brokers to stand up and say, wait a minute, we never intended to cover this sort of event.”

Risk Disclosures and Exemption ClausesThe market involves risks, and investment needs caution. This article does not constitute personal investment advice, nor does it take into account the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their circumstances. Any investment based on this is at the user's own risk.