How to navigate a safe AI transformation

Artificial intelligence is no longer a future trend. It is already reshaping the global economy at an unprecedented pace.

Adoption has moved beyond experimentation to widespread use across industries. Organizations are changing how they operate, make decisions, and serve customers.

For insurers, this shift is especially relevant. We are not just observers of AI. We use it across our operations while also underwriting the risks it creates for businesses and society.

This dual role puts the insurance industry in a unique position to help ensure that AI delivers value safely, responsibly, and at scale.

From opportunity to responsibility

The benefits of AI are already visible. Across insurance, it improves customer experience, accelerates claims processing, and expands access to expertise. For example, AI-powered systems can reduce handling times and increase customer satisfaction, while allowing experts to focus on complex cases.

But opportunity also brings responsibility.

AI systems influence decisions that matter, including coverage, claims, pricing, and customer interactions. This makes trust, governance, and accountability essential.

At Zurich, this means embedding governance at every stage of development. Systems are designed to be transparent, explainable, and continuously monitored. Clear ownership and oversight are in place. High-impact applications are subject to stricter controls, including risk assessments, bias testing, and structured incident management.

For decisions with significant consequences, a human remains involved to ensure judgment, empathy, and accountability.

The goal is not to eliminate risk. It is to manage it effectively. .

Understanding the AI risk landscape

To manage AI safely, we must first understand its risks.

Insurance provides a valuable perspective. Experience with areas such as cyber risk shows that new technologies rarely create entirely new risks. Instead, they reshape and amplify existing ones while introducing new, harder-to-measure exposures.

The AI risk landscape can be viewed in three layers:

1. Known and manageable risks
   
   These are familiar challenges that AI intensifies:
   
   • Bias and discrimination in decision-making
   • Incorrect or unsafe outputs
   • Data privacy and security risks
   • Intellectual property concerns
   These risks are well understood and can be mitigated through strong governance and controls.
2. Emerging risks
   
   More complex challenges are now becoming clearer:
   • Limited explainability of AI decisions
   • Unclear accountability across complex supply chains
   • AI-enabled cyber threats
   Here, traditional frameworks need to adapt.
3. Systemic risks
   
   At the frontier are risks that could affect entire systems:
   • Correlated failures across similar AI systems
   • Concentration risk among a small number of providers
   • Cascading effects across interconnected systems
   These risks are the most difficult to measure and potentially the most impactful.

Why insurability matters

Insurance depends on a simple principle: risks must be understood and attributable to be insurable.

AI challenges this principle.

As systems become more complex, especially with autonomous and agentic AI, it becomes harder to understand how decisions are made and who is responsible. In multi-layered ecosystems involving developers, platforms, and users, accountability can become unclear.

Without clear responsibility, risks are harder to price and insure.

This is critical because insurance enables innovation. Without mechanisms to transfer risk, adoption slows and trust declines.

Building trust in the age of AI

What does it take to move forward safely? Three priorities stand out:

1. Transparency and accountability
   
   People need to know when AI influences decisions. They must be able to understand, question, and challenge outcomes.
   
   Clear rules on liability are equally important. When responsibility is well defined, risks can be priced, insured, and managed.
2. Proportionate governance
   
   Not all AI systems carry the same level of risk. Governance frameworks should reflect this.
   
   High-impact applications require strong oversight, human involvement, and rigorous testing. Lower-risk uses should not face unnecessary complexity.
   
   Systemic risks such as concentration or cascading failures also need collective solutions, including international coordination and shared standards.
3. Investing in people
   
   AI will reshape work, and this transition must be actively managed.
   
   Reskilling, education, and workforce support are essential—not only to mitigate risk but also to unlock AI’s full potential.
   
   Human skills such as judgment, ethics, and creativity remain essential and must be preserved.

A shared responsibility

AI transformation is already underway. The question is not whether it will happen, but how.

The choices we make today on governance, accountability, and collaboration will shape whether AI becomes a source of trust and progress—or fragmentation and risk.

Insurance has always played a key role in managing uncertainty and enabling progress. The same applies here.

The goal is not zero risk.

It is risk that is understood, governed, and proportionate—creating the foundation for AI to scale safely, responsibly, and with confidence.