When Your AI Calls The Insurer’s AI, Who Owns The Mistake?

Estimated reading time: 7 minutes

Consumer AI agents have started calling insurers to shop for coverage. In a conversation with Cyber Insurance News, Liberate co-founder and CEO Amrish Singh said a licensed human still signs. Who pays when the machine gets coverage wrong remains unsettled.

The conversation began with a canceled flight.

A major international airline canceled a flight I had booked. An automated email confirmed a rebooking on the same flight a day later. The airline’s app showed a different flight, five hours earlier.

After 40 minutes on hold, a representative declined responsibility for the email. “AI sends those emails,” she said.

Singh went straight to the point. “But wait, the AI is working on behalf of them, right?” he said. “I 100% would believe that this is part of their errors and omissions risk.”

Singh co-founded Liberate. Its AI agents answer calls, emails, and texts for property and casualty insurers and agencies. Liberate says it handles more than 3.5 million voice transactions a month for more than 70 carriers and brokers. Singh has spent almost 18 years in insurance.

Consumer AI Agents Start Calling Insurers

Meta began rolling out Muse, its personal AI agent, in the US on 8 September, Reuters reported. RuntimeWire reported that Meta is privately testing Muse calls to verified US business numbers.

Liberate launched AI Intercept on 28 September. It screens inbound calls to carriers and agencies and flags consumer AI agents.

AI callers remain a small share of traffic. “I wouldn’t say that we have a large portion,” Singh said. “We’ve seen about a 3x increase in the sheer number of calls coming in. But the volume of calls is still low right now.”

“Interestingly, and I don’t know why, we see it in spikes,” he said.

For now, the agents mostly shop for quotes, Singh said. Service, refund and cancellation requests remain rare. “The rate of change is what we are more keeping an eye on,” he said.

Insurify blocked Muse from its comparison platform on 23 September, Insurance Journal reported. Insurify said Muse’s results dropped limits, deductibles, discounts, and eligibility details. “A quote without its context is not a fair comparison,” said co-CEO Giorgos Zacharia.

A Licensed Human Still Signs

I asked who is the licensed producer of record when Liberate’s AI deals with a consumer’s AI. Singh said Liberate’s AI does not quote.

“Because of the way the laws are right now, our AI is not licensed,” he said. “It’s not allowed to quote consumers.”

Liberate’s AI handles data entry and collection. It then hands off to a licensed agent. “It’s up to the licensed agent at the end to verify that information before they bind a policy,” Singh said.

The agent gets a warning first. “We tell the human in the loop, the agent, listen, you’re talking to an AI,” Singh said. “So be prepared to make a determination based on the fact that it’s an AI calling in.”

Carriers and agencies set their own routing rules. An AI responds only when the system clears a confidence threshold. “Typically it’s in the higher 90s,” he said. Otherwise, a human takes the call.

“It’s customizable,” he said. “Not a hard binary decision.”

AI Voice Fraud And The Identity Check

I asked whether Intercept can tell a legitimate consumer agent from a fraudster’s voice agent. Singh treats those as two problems. Detecting an AI voice is one signal. Fraud is another.

“The act of an AI attempting to make a fraudulent action is actually slightly different,” he said. “You could hire hundreds of humans to conduct fraudulent calls, and they are as scary as AI making fraudulent calls.”

Liberate runs identity verification separately, Singh said. It checks on whose behalf a caller is acting and how often they have called. “We can stop that instantly,” he said.

Criminals already sell voice agents of their own. Halcyon’s Ransomware Research Center found an AI-powered outbound call center advertised on criminal forums. It places up to 120 calls at once, in up to 25 languages.

Consumer AI Agents And The Cost Of Shopping

Singh expects every business to need an AI front door. “The way we used to have a website for a digital storefront, you need to have an agent responding to it from an agent storefront,” he said.

Shopping around now costs a consumer almost nothing, Singh said. “I can go to my AI agent Muse and say, hey, call every imaginable insurance carrier you know of, and here’s my risk profile. Get me the best rate imaginable.”

Insurers have long been “relying on the inefficiencies of the human time to dictate how much they’re going to drive the cost on the insurance company,” he said.

AI Liability Remains Unsettled

“If a coverage turns out to be wrong, like who owns that?” Singh said. “The industry has to figure this out.”

“We don’t have an answer for it, but it’s a thing that we are all worried about,” he said.

Outside insurance, one tribunal has ruled. In 2024, British Columbia’s Civil Resolution Tribunal held Air Canada liable for its website chatbot’s wrong fare advice. The airline had argued the chatbot was responsible for its own actions, the ABA’s Business Law Today reported.

RAND’s study of insurance filings found the largest group of carriers neither affirms nor excludes AI.

Optimism With Conditions

Singh calls himself an eternal optimist. “The technology innovation is equally available to both parties.”

His worry is other vendors. “I fear the maybe misplaced optimism sometimes, which is like cool technology, I want to disrupt the space,” he said.

“AI can’t make mistakes in the P&C insurance industry because the implications are dramatic,” Singh said. “What do you do when it makes a mistake? How do you track it?”

My airline dispute ended with AI on both sides. I used a consumer AI tool to draft a compensation claim under European passenger rules. The claim succeeded.

FAQ – Consumer AI Agents

Are consumer AI agents calling insurers?

Yes. Liberate CEO Amrish Singh said AI-placed calls to the carriers and agencies it serves have roughly tripled, though volume remains low. Most calls are shopping for quotes.

Can an insurer’s AI quote or bind a policy for a consumer’s AI?

Not at Liberate. Singh said its AI is not licensed and does not quote. It collects information and hands off to a licensed agent, who verifies it before binding.

Does the licensed agent know an AI is calling?

Yes. Singh said Liberate tells the agent the caller is an AI so they can make their determination with that in mind.

Can AI detection stop voice fraud?

Singh treats AI detection as one signal. Liberate runs identity verification separately, checking on whose behalf a caller acts and how often they have called.

Who is liable if an AI gets coverage wrong?

That remains unsettled. Singh said the industry has to figure it out. In 2024 a Canadian tribunal held Air Canada liable for its chatbot’s wrong advice.

What is Muse?

Meta’s personal AI agent, which began rolling out in the US on 8 September 2026. Meta is testing calls to verified US businesses. Insurify blocked Muse from its comparison platform.

Related Cyber Insurance Posts

• OT Cyberattacks Cost About $1 Million. Fewer Than One In Four Listed An Insurance Claim.
• Nobody Owns The Password. Passpack CEO Chris Skipworth Explains The Credential Gap Sitting Inside Your Next Renewal – PODCAST(Opens in a new browser tab)
• Deepfake Fraud Surges 1,300% According to Pindrop’s 2025 Report(Opens in a new browser tab)
• Insurers Focus on Cyber Coverage to Boost Auto and Property Profits(Opens in a new browser tab)
• Identity Risk Shifts As AI Drives New Wave Of Impersonation Attacks(Opens in a new browser tab)