Mumbai | New Delhi: What if an AI agent goes rogue and deletes the entire database of your company? Or, if your chatbot lies to the customer? These are the questions chief information security officers at leading companies are asking and the reasons they are increasingly turning towards cyber insurance to safeguard against “AI liability”.

Private sector insurer HDFC ERGO said it has seen a 25-30% increase in enquiries from companies buying cyber insurance for the first time. While it remains a specialised line, its relevance is growing as organisations seek structured risk-transfer solutions, it said.

Premiums for high-risk sectors have grown by 15-25% year-on-year globally, while in India, the increase is up to 10%, according to Prudent Insurance Brokers.

Underwriting models are also evolving. Insurers have started assessing companies’ AI-exposure, governance and even dependence of AI coding tools to include in underwriting frameworks, especially where the firms have declared AI-related revenue.

“This shift is increasingly reflected in underwriting practices, policy wordings and risk assessment frameworks, with greater emphasis on clarity around coverage, exclusions and responsibilities,” said HDFC ERGO chief underwriting officer Niraj Naik.

“Insurers are also examining AI development and deployment practices, including standards and processes for critical areas like data management, robustness and stability, explainability and fairness, reproducibility and auditability,” said Tanuj Gulani, president, Prudent Insurance Brokers.

Deloitte forecasts the global AI insurance market to be a $4.8 billion annual opportunity by 2032, growing at an 80% annual rate.

However, AI risks are shaking up traditional actuarial models, because they do not have historical loss data and stable risk patterns, which are critical for accurate risk pricing and prediction models.

Ritesh Thosani, cyber practice leader at Marsh India Insurance Brokers, said industry stakeholders including regulators are discussing AI threat vectors such as automated discovering vulnerabilities or malicious code manipulation to efficiently price these risks.

“Movement in pricing is usually driven by claims and losses incurred by insurance companies, which we are yet to see. Therefore, the insurance market, especially in India, remains competitive. As more data becomes available, we may see changes in how generative AI risks are underwritten and priced,” he said.

A large part of AI risks stem from the fact that organisations have started deploying AI without fully understanding the guardrails, said Amol Bhat, partner, Cybersecurity at PwC India.

“AI by itself may not add more risk premium, what will, is whether entities are using the right guardrails in using AI and are they using this for both operations and defence in depth or both,” he said.