The rapid advancements in artificial intelligence (AI) have ushered in a new era for businesses, allowing them to tap into the power of autonomous “agents” that are designed to enhance efficiency and drive revenue growth. However, as more companies place their trust in AI, a new layer of risk has emerged that highlights the uncertainties involved in this technological leap. Some insurance firms are beginning to respond by cautiously stepping up to the plate to offer coverage for potential missteps, while others remain hesitant due to the inherent complexities involved.

Phil Dawson, head of AI policy and partnerships at the specialist insurer Armilla, emphasizes that the key motive behind deploying advanced AI is to significantly reduce human oversight in crucial decision-making processes. This growing trend of utilizing “agentic AI”—programs that operate independently—has led to a significant reshaping of the workplace, with organizations trimming down their workforce as they increasingly rely on automated systems. Yet, this development presents fundamental challenges to traditional insurance frameworks.

The crux of the disruption lies in how current insurance policies are structured. Traditionally, insurance firms have largely accounted for AI-related liability risks within what is known as “silent coverage.” This practice allows companies to operate under the assumption that certain liabilities are implicitly covered, but experts like Sonal Madhok and law professor Anat Lior point out that such a passive approach risks leaving businesses exposed. Their research, published via brokerage firm Willis Towers Watson, suggests that the next iteration of insurance will inherently need to address AI-specific coverage clearly.

Madhok and Lior anticipate a shift away from silent coverage toward more explicit policies addressing AI risks. As firms grapple with the realities of AI-induced errors—phenomena like “hallucinations,” where systems present fabricated information with confidence—insurers are adapting their policies to either include or exclude such risks. Jonathan Mitchell, head of the financial sector practice at brokerage firm Founder Shield, emphasizes the transition from a “wait-and-see” approach to a more proactive stance regarding AI liabilities.

For instance, some standard insurance policies have already introduced “absolute AI exclusion” clauses that explicitly deny coverage for AI-related incidents. This evolution reflects a deepening understanding of the risk landscape shaped by these technologies, pushing both businesses and insurers to reassess what constitutes adequate coverage in this new paradigm.

Illustrative of the risks at play is a case involving a commercial real estate firm that sought coverage for its AI agent as if it were a regular employee. This exemplifies the transformative nature of AI in business operations but also highlights the nuances and challenges faced by insurers as they develop strategies for risk management and coverage determination in such evolving contexts.

Moving to incorporate specific AI-related concerns, companies like Founder Shield are now offering policies designed to cover losses caused by issues such as AI malfunctions and hallucinations. These revisions can extend beyond just network-related issues to cover tangible impacts—potentially addressing scenarios like an AI mistakenly ordering excessive inventory, which could be detrimental for businesses.

Despite these advancements, insurers like Armilla remain cautious, vetting AI models for vulnerabilities before extending coverage. While they can reject certain high-risk scenarios, their focus on compliance with international standards indicates a thorough and responsible approach to underwriting in this uncertain landscape. The firm selectively avoids providing coverage for medical diagnostics and applications focusing on mental health, recognizing the heightened level of risk associated with these areas.

Meanwhile, Munich Re, a global giant in both insurance and reinsurance, has begun offering coverage that caters to businesses creating and utilizing AI models. Their head of AI insurance, Michael von Gablenz, acknowledges the inherent unpredictability in model behavior, emphasizing that statistical models come with their own uncertainties—a recognition that informs how risks are handled.

As the landscape surrounding AI continues to evolve, the dialogue between insurers and businesses promises to reshape the understanding of liabilities, potentially leading to an era where policies are tailored to address the specific needs arising from AI technology. This ongoing dialogue is crucial to ensuring that businesses can innovate with confidence while protecting themselves against the unforeseen repercussions of their automated decisions.