A British Columbia man named Jake Moffatt booked a discounted bereavement airfare after Air Canada's chatbot told him he qualified. Air Canada later refused to honor it. The airline's defense, submitted in actual court filings, was that its chatbot was a "separate legal entity" responsible for its own actions, so Air Canada couldn't be held accountable for what the bot said.
The Civil Resolution Tribunal rejected that argument. Air Canada was liable. The chatbot was the airline's agent. Its mistakes were the airline's mistakes.
That case settled in early 2024 and didn't get the attention it deserved in insurance circles. Because the question it raised, who's actually on the hook when your AI agent makes a mistake, still doesn't have a clean answer in most commercial insurance policies. And Illinois businesses are deploying AI agents right now without knowing whether their coverage reaches the exposure.
What the Air Canada case actually established
Moffatt v. Air Canada isn't complicated. You can't deploy an AI system to interact with customers, give it authority to make representations on your behalf, and then disclaim responsibility when it gets something wrong. Courts won't accept that framing.
The airline set up the chatbot. It trained the chatbot. It pointed customers to it. When the chatbot gave incorrect information about bereavement fares, that was Air Canada's problem to fix. Not the chatbot's. Not some notional entity separate from the company.
Your AI agent acts on your behalf. When it acts badly, that's on you. And that's where the insurance problem begins.
The three categories that drive commercial insurance
Commercial insurance is built around three types of legal responsibility: products, services, and employees. Almost every major policy form keys off one of them.
Products get covered by product liability. A physical thing your company made causes harm, product liability responds.
Services get covered by professional liability, also called errors and omissions (E&O). You gave advice or performed a service that caused a client financial harm, E&O is the policy that matters.
Employees get covered through employers liability, workers compensation, and the vicarious liability principles embedded in general liability policy forms. When someone working for you causes harm while acting within the scope of their job, you're generally on the hook, and that exposure flows through your commercial program.
An AI agent doesn't fit cleanly into any of those categories. It's not a physical product. It's not a human professional. It's not an employee. That's the gap.
General liability: not built for this
Your general liability (GL) policy covers bodily injury and property damage. That's the core of what it does. A customer slips on your floor, GL responds. A contractor damages a client's building, GL responds.
What GL doesn't cover well is economic harm that doesn't trace back to something physical. If your AI agent gives a client the wrong pricing information, books a vendor at the wrong rate, or generates a contract term that ends up costing someone money, that's financial harm. Not physical harm. Most GL policies have explicit exclusions for economic losses that don't stem from bodily injury or property damage.
Some GL forms have personal and advertising injury provisions that could theoretically reach certain AI mistakes, like defamatory statements or false advertising claims. But that's not a reliable path. It's fact-specific and policy-specific, and "might work" isn't the same as "will work when you actually need it."
E&O: a closer fit, with real gaps
Professional liability is more naturally suited to AI-related mistakes. E&O covers you when an error or omission in your professional services causes a client financial harm.
But it has limits that matter here.
E&O is designed around professional judgment, specifically human professional judgment. When your AI agent produces a recommendation, analysis, or output that turns out to be wrong, whether that qualifies as a "professional service" under the policy is an unsettled question. Some underwriters are comfortable covering it. Others aren't. When an underwriter isn't comfortable, that ambiguity shows up as an exclusion or a coverage dispute at the worst possible time.
E&O also typically covers claims from clients, people you had a direct professional relationship with. If your AI agent's mistake harms a third party who wasn't your client, your E&O may not respond at all.
And this is increasingly significant: E&O policies are starting to add AI exclusions explicitly. ISO, which develops the standard policy forms most carriers use, filed AI-specific endorsements in January 2026. If your policy was renewed or newly issued after those changes took effect and your carrier adopted them, AI-generated outputs may be outside your coverage entirely. You need to know whether your policy has one of these endorsements before you find out at claim time.
Product liability: only part of the picture
If your business actually builds AI software, product liability starts to become relevant. But software has always occupied a murky space in product liability law. Courts have historically split on whether software is a "product" or a "service," and cloud-delivered software tends to get treated as a service, which means product liability doesn't reach it.
Product liability is also fundamentally designed around physical harm. When a defective AI output causes financial loss rather than bodily injury, product liability often doesn't get there.
There's also a vendor dimension worth understanding. If you're an Illinois manufacturer or contractor using an AI tool licensed from a third party and that tool makes a costly mistake, product liability points at the vendor. But you may still face a claim from whoever absorbed the loss, and whether your own coverage responds to that claim is a separate question from whether the vendor is ultimately at fault.
Vicarious liability and the "it was the AI" defense
Back to Air Canada. The court's reasoning reflects a basic legal principle that's been settled for over a century: when someone acts on your behalf with your authority, you're responsible for what they do.
This principle, vicarious liability, applies to employees. If your employee causes harm while acting within the scope of their job, you're generally on the hook. That liability flows through to your GL policy through the employer's liability provisions and the broader vicarious liability framework baked into those forms.
But AI agents aren't employees. They don't have legal personhood. They can't be personally liable. Workers compensation doesn't cover them. The employer provisions in your GL weren't written with AI agents in mind.
You're still responsible for what your AI agent does, because you deployed it, authorized it, and pointed customers or vendors at it. Courts will hold you accountable, as the Air Canada tribunal did. But your insurance program may not have caught up to that reality.
What this looks like in practice
These aren't speculative scenarios. They're the kinds of situations Illinois businesses are already navigating.
A Chicago-area property management company deploys an AI agent to handle tenant inquiries. The agent tells a tenant their lease allows subletting when it doesn't. The tenant sublets, violates the lease, and the resulting dispute costs the property company $48,000 to resolve. GL doesn't respond because it's economic harm with no physical component. The E&O policy has a clause that might reach it, but the claim is in dispute with the carrier.
A DuPage County accounting firm uses an AI tax preparation assistant. The tool misreads a new Illinois tax provision. Three clients file late and face combined penalties of $41,000. The firm gets E&O claims. Because the policy predates the January 2026 ISO exclusion filings, coverage applies. But renewing that same policy now means facing a form with AI exclusions baked in.
A west suburban manufacturer uses an AI procurement agent to place purchase orders. The agent applies a pricing formula incorrectly and commits the company to $175,000 in materials at the wrong contract price. Pure economic harm. No bodily injury. No property damage. Six weeks of legal analysis later, nobody's commercial insurance program covers it cleanly.
What Illinois businesses should audit right now
If your business runs AI tools that interact with customers, vendors, or financial systems in any decision-making capacity, four things are worth reviewing with your agent.
Your GL policy's economic harm coverage. Does your GL cover financial losses that don't stem from bodily injury or property damage? Most don't, but knowing exactly what you have is the starting point for identifying the gaps.
AI exclusions in your E&O or professional liability policy. ISO's January 2026 endorsements added significant AI-related language to standard policy forms. Several carriers have filed their own AI exclusions with state regulators, including in Illinois. Your policy may have them already. Ask your agent to pull the endorsement schedule and identify any AI-related language before you need to rely on the coverage.
Your vendor contracts. If you're using a third-party AI tool, your agreement with that vendor probably has indemnification language. But "probably has indemnification" and "definitely covers this specific scenario" are different things. The clause may have carve-outs, limits, or conditions that matter when an actual claim lands.
Whether your program needs a specialty AI liability endorsement or policy. A handful of insurers are actively writing AI liability coverage now. Some are designed for companies that build AI tools, some for companies that use them, and that distinction matters in how the coverage is structured and what it actually reaches.
The honest position for Illinois businesses in 2026
Standard commercial insurance wasn't built with AI agents in mind. The categories were developed long before these tools existed, and the policy forms are updating slowly. That gap between what Illinois businesses are deploying and what their insurance programs actually cover is real.
You're running tools that make decisions on your behalf, interact with customers, and create binding commitments. Courts will hold you accountable for those decisions, the same way Air Canada was held accountable for what its chatbot said to Jake Moffatt. Whether your insurance responds the way you expect is a separate question, and it's one worth getting an honest answer to now rather than after a claim.
If your Illinois business uses AI agents in customer service, procurement, financial operations, or any client-facing role, reach out to Jack Ray directly. He works with commercial clients across the state and can walk through your current program, identify where AI exposure isn't covered, and connect you with markets actively writing coverage for it in 2026.
Email: jray@lakeshoreriskadvisors.com