AI-related litigation in the United States has surged 978% between 2020 and 2025, exceeding 700 cumulative lawsuits. Meanwhile, 95.2% of the $1.63 billion raised by InsurTechs in the first quarter of 2026 went to AI-focused companies. The market is chasing a risk it doesn’t yet know how to price.

This gap is not incidental. It says something precise about the next wave of automation: it won’t be engineers who decide who deploys humanoid robots, autonomous vehicles, or AI agents in warehouses and hospitals. It will be actuaries.

The Essentials

  • AI-related litigation in the United States jumped 978% between 2020 and 2025, exceeding 700 cumulative lawsuits, according to the “Smart Systems, Blind Spots” report by Gallagher Re, MIT, and Testudo (March 2026).
  • China launched its first insurance products dedicated to humanoid robots in September 2025, while Western insurers began adding AI exclusions to their standard policies.
  • This divergence creates an access divide: large groups can self-insure; SMEs remain locked out due to lack of available coverage.
  • 95.2% of the $1.63 billion raised by InsurTechs in Q1 2026 went to AI-focused companies, a sign that the sector recognizes the urgency without yet having the answers (Gallagher Re, Q1 2026 Global InsurTech Report).

When Technology Moves Faster Than Accountability

Humanoid robots are no longer lab projects. Figure AI is deploying its units in BMW factories in South Carolina. Agility Robotics is supplying its Digit robots to Amazon. Unitree is selling its models at prices below $20,000, making the technology accessible to mid-sized companies. Goldman Sachs projections estimate the global market at $38 billion by 2035.

But accelerating deployment means accepting incidents. A robot that falls on an employee, an autonomous vehicle that injures a pedestrian, an AI agent that makes a wrong medical decision: who pays? Under current legal frameworks, the answer is rarely clear. And when the answer is unclear, insurers exclude rather than cover.

That’s precisely what’s happening. General liability policies in the United States and Europe are beginning to incorporate specific exclusion clauses for “autonomous systems” and “algorithmic decisions” starting in 2025, with standardized Verisk/ISO forms effective as of January 2026. These exclusions, designed to limit insurers’ exposure to unmodeled risk, create a coverage gap precisely where deployment is accelerating. An SME wanting to install an autonomous robotic arm in its warehouse discovers it’s not covered by its existing policy, and that no dedicated product yet exists on the Western market.

China Opens the Market That the West Still Hesitates to Create

In September 2025, People’s Insurance Company of China (PICC) and Ping An launched the first insurance products specifically designed for humanoid robots. These policies cover bodily injury to third parties, mechanical failures, and incidents related to autonomous decisions. They are backed by real-time telemetry data: the robot continuously transmits its operating parameters, allowing the insurer to adjust premiums based on the machine’s actual behavior.

This model is not incidental. It assumes shared data infrastructure between manufacturer, operator, and insurer. It also assumes regulatory tolerance for data collection. Both conditions are more easily met in China than in Europe, where GDPR and fragmentation of industrial actors complicate this type of architecture.

China’s advantage is not only technological; it is institutional. Beijing understood before others that insurance coverage is a condition for large-scale deployment, not an administrative issue to be resolved after the fact. The government explicitly included the development of insurance products dedicated to humanoid robots in its industrial roadmap published in late 2024. The signal is clear: insurance is treated as infrastructure, on par with electrical grids or 5G connectivity.

In Europe, the regulatory framework applicable to AI systems now rests on the AI Act, which entered into force in 2024, and on the new Product Liability Directive (Directive (EU) 2024/2853, published in November 2024). These texts remain focused on risk classification, burden of proof, and victims’ right to compensation. They create no mandatory insurance mechanism nor pricing framework. European insurers are waiting for claims data that will only arrive if deployment begins. Deployment won’t begin without coverage. The loop is closed.

An Access Divide Between Large Groups and SMEs

The absence of standardized products does not block everyone equally. Large industrial groups have an option that SMEs don’t: self-insurance, also called a captive insurance company. Amazon, BMW, or Foxconn can create internal entities that assume the risk directly, provisioning potential losses on their balance sheets. It’s expensive, but it’s legal and manageable for companies with balance sheets of tens of billions of dollars.

For a 200-person manufacturing SME wanting to automate a production line, self-insurance is not an option. It needs a market product. And that product doesn’t yet exist in most Western markets. Result: large groups deploy, SMEs wait. This isn’t a market mechanism favoring the most efficient; it’s an access mechanism favoring the best capitalized.

This divide has concrete consequences for the geography of automated transition. Regions whose industrial fabric is dominated by SMEs—as is the case for much of continental Europe—risk structural delays not because technologies are unavailable or too expensive, but because the insurance framework to deploy them legally doesn’t exist. It’s the same mechanism that, in other sectors, long delayed innovation adoption in economies that were otherwise rich but institutionally behind.

We’re seeing a comparable phenomenon in labor regulation in the AI era: jurisdictions that move first define the rules of the game for others, often advantaging existing actors.

InsurTechs Chasing Risk They Don’t Yet Understand

The $1.63 billion raised by InsurTechs in the first quarter of 2026 isn’t going toward mature products. It’s financing mostly companies attempting to build actuarial models that allow pricing of AI and robotics risk. This is a market in formation, not an established one.

Several approaches are emerging. One family of startups bets on telemetry data, the same model as China: collecting real-time operating parameters of autonomous systems to calculate dynamic premiums. It’s technically elegant but commercially tricky, as it assumes operators accept sharing data on their industrial processes with a third party.

A second family is working on “risk sandboxes”: controlled testing environments where autonomous systems accumulate simulated claims data before real deployment. The idea is to artificially create the statistical base missing for pricing risk. It’s a serious path, but one that takes time and whose results depend on the quality of simulated scenarios.

A third family, more modest in scope, tackles the problem of distributed liability. When an incident involves a humanoid robot, multiple actors can be implicated: the robot manufacturer, the autonomy software developer, the integrator who configured the machine, the operator who deployed it. Determining the liability chain is as much a legal problem as a technical one. Startups like Bryte Insurance in the United States or Elco in Europe are working on “value chain” policies that cover all actors involved in a deployment, with responsibility allocation defined contractually in advance.

These initiatives are real. But they remain experimental, and the gap between the speed of robotic deployment and the speed of insurance market maturation is widening. AI produces, prepared organizations capture: the same logic applies here to financial institutions—insurers building their actuarial expertise now will capture the market when it tips.

What Law Doesn’t Yet Resolve

The “Smart Systems, Blind Spots” report by Gallagher Re, produced with MIT and Testudo in March 2026, points to a fundamental legal problem that insurance policies alone cannot solve: the absence of a clear legal category for autonomous decisions.

In classical tort law, you must identify a fault, an author, and a causal link. When a humanoid robot makes an autonomous decision that causes damage, no human has committed a fault in the traditional sense. The engineer who programmed the machine’s general behavior did not decide this specific act. The operator who deployed the robot didn’t order the incriminated action either. The robot itself has no legal personality.

This gap is not theoretical. It concretely determines the difficulty in handling the 700 lawsuits already filed in the United States and insurers’ reluctance to offer coverage. If the insurer cannot be certain it will be subrogated in the victim’s rights to recover from the responsible party, it hesitates to pay. And if it hesitates to pay, it hesitates to cover.

Several jurisdictions are tackling this problem. The European Parliament debated in 2024 an “electronic personality” for AI systems, an idea abandoned in the final version of the texts but which will resurface. The state of Delaware in the United States is experimenting with a registry of autonomous systems that assigns a legal identifier to each commercially deployed AI agent, without legal personality but with tracing of the responsibility chain. It’s a pragmatic approach that sidesteps the philosophical debate to meet operational needs.

Maritime law offers perhaps the best precedent. For centuries, ships have caused damage without always identifying a direct human responsible party. The solution wasn’t to personify ships, but to create strict liability regimes (the owner is liable for damages caused by their ship, regardless of personal fault) backed by mandatory insurance. The same logic applied to autonomous systems would break the current deadlock.

The Next Bottleneck to Break Through

The history of technological diffusion is also the history of creating institutions that enable that diffusion. Mass automobiles had to wait for mandatory insurance to become a universal good. Commercial aviation had to wait for certification standards and accident investigation protocols to gain public trust. Domestic electricity had to wait for safety standards and mandatory inspections. In each case, the technology was ready before the institutional framework was.

Humanoid robots and autonomous systems are probably in the same phase. Technical performance is advancing quickly. Boston Dynamics, Figure AI, Unitree, and Agility Robotics produce machines capable of executing complex tasks in unstructured environments. Production costs are falling. Use cases are clarifying. What’s missing is the institutional framework that allows an ordinary company, not just Amazon or BMW, to deploy these machines with visibility on its liability and coverage of residual risks.

China has the advantage of moving first and treating insurance as national infrastructure. Europe and the United States have the advantage of deeper insurance markets and more developed legal traditions regarding tort liability. The question is not whether these markets will create suitable products, but how quickly they will. And whether that speed will be fast enough for Western SMEs to participate in the deployment wave, or whether they’ll watch it pass from the stands.

Sources

  1. Gallagher Re — Q1 2026 Global InsurTech Report: https://www.ajg.com/gallagherre/news-and-insights/global-insurtech-report-for-q1-2026/
  2. Gallagher Re, MIT, Testudo — “Smart Systems, Blind Spots”, March 2026 (available via Gallagher Re): https://www.ajg.com/gallagherre/news-and-insights/smart-systems-blind-spots-rethinking-insurance-for-the-ai-era/
  3. European Parliament — AI Liability Directive (Legislative Train): https://www.europarl.europa.eu/legislative-train/theme-a-europe-fit-for-the-digital-age/file-ai-liability-directive
  4. Goldman Sachs — Humanoid Robot Market Projections 2035 (Equity Research report, 2024)
  5. People’s Insurance Company of China (PICC) — Announcement of humanoid robot insurance products, September 2025
  6. State of Delaware — Autonomous Systems Registry Experimentation, 2024-2025
  7. China Daily — Insurance policy for humanoid robots, December 2025: https://www.chinadaily.com.cn/a/202512/12/WS693b6d05a310d6866eb2e404.html
  8. Figure AI — F.02 Contributed to the Production of 30,000 Cars at BMW: https://www.figure.ai/news/production-at-bmw
  9. Unitree Shop — Humanoid Robot Catalog: https://shop.unitree.com/collections/humanoid-robot
  10. Business Wire — Agility Robotics Broadens Relationship with Amazon: https://www.businesswire.com/news/home/20231024174668/en/Agility-Robotics-Broadens-Relationship-with-Amazon