Beyond the Hype: 5 Hidden Liabilities of Using Generative AI in Your Business

The 5 Heavy-Hitter Liabilities

Reading about cases such as the Air Canada AI chatbot offering a ticket discount that didn’t actually exist, it’s easy to understand why some companies might run the other direction when generative AI enters the conversation. However, these liabilities shouldn’t discourage leaders. Instead, the best way to avoid a problem is to acknowledge it, study it in depth, and create strategies to mitigate risks.

1. Generative AI Errors & Misrepresentation

The Air Canada example perfectly demonstrates AI errors and misrepresentations. Simply put, a passenger spoke to a chatbot about receiving a bereavement fare for his grandmother’s funeral after going on his flight. Unfortunately, the chatbot assured him that this was possible despite Air Canada’s policy that the discount would only apply before the flight took place.

Although it was the chatbot that gave the wrong information, Air Canada was still held liable for this mistake and was made to pay for the passenger’s refund, plus damages and legal fees. The tribunal member said, “It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.”

Ultimately, when a customer-facing chatbot speaks to the public, its words carry legal weight. For customers, this chatbot is considered an extension of the corporation, and the information it provides is taken as an accurate company policy.

2. IP Infringement and Defamation

Generative AI is no stranger to marketing teams these days. More often than not, online ads, social media posts, and even billboards are all crafted using this technology. And, although it’s a great shortcut for creative teams, it’s easy to forget that most LLMs are trained on copyrighted data, which means unintentional plagiarism or hallucinated false statements about competitors that make the final draft are never out of the question. In turn, a massive legal exposure is created.

For example, suppose there’s an artist with a very distinct painting technique that makes their art instantly recognizable. Unbeknownst to a company’s marketing team, the visual prompt they ask their AI returns quite a similar art style to the painter, who is made aware of the campaign by social media followers. What follows is a costly copyright lawsuit, as the marketing team technically copied the painter’s trademark, even if unintentionally.

The lack of intent to infringe is not a valid legal defense in copyright law, which means the team is entirely liable for such a mishap. If the company publishes any of this sort, it is directly responsible and can be sued.

3. Unauthorized Data Disclosure

In an era where data is the undisputed currency of modern business, rigorous data governance and elite cybersecurity have become industry standards. That is exactly why traditional software relies so heavily on data silos and strict access controls. Whether dealing with sensitive customer information or proprietary company data, these barriers ensure that only authorized users can view, manipulate, or extract the precise information they need to do their jobs.

However, this isn’t the case for generative AI. Popular LLMs are inherently designed to process vast amounts of data, all pooled into one lake or database to continue training the model. While the tool becomes smarter with all of this information, it is also capable of extracting data the same way it is fed to it, unless the proper safeguards are put into place. What this means is that, if any sensitive information is given to a generative AI tool, it can also disclose it just as easily, possibly cross-contaminating private user data by accident.

If a hotel concierge uses an internal-facing chatbot to ask for specific guest notes to give staff, and the AI discloses an adjacent guest’s highly sensitive medical accommodation by error, a lawsuit could be underway if this information spreads.

Very casually, an automation tool has created a data privacy nightmare for the hotel chain by triggering compliance and regulatory violations, especially in places like Europe with GDPR and California with CCPA rules. To put these figures into perspective, an average GDPR fine between 2018 and 2025 cost companies €2.36 million, or $2.73 million.

4. Bodily Injury

AI is entirely digital, but it can be used to dictate physical instructions across many industries, including healthcare. In fact, 80% of US healthcare leaders reported that they’ve already deployed their first AI use cases to end users. This isn’t without its risks—in healthtech, workplace safety, or wellness, a faulty algorithmic recommendation can cause physical harm.

For instance, many healthcare providers could be adopting AI to help them come up with patient prescriptions more quickly than they would manually. Let’s say that, on one occasion, the generative AI happened to miss a critical, well-documented drug interaction that the physician failed to double-check before sending this information to the patient. This new prescription leads the patient to be hospitalized upon taking the recommended dose, opening the door for a costly legal battle.

With today’s wide range of use cases, and though it seems far-fetched, AI decisions can easily leak into physical outcomes that end up in real-world harm and catastrophic bodily injury liability claims.

5. Property Damage

Similarly to AI’s newfound impact in physical scenarios, relegating physical workflow decisions to this technology in areas like logistics and construction can bring property damage liabilities. Even backend, B2B SaaS, or scheduling tools can destroy physical assets if they are trusted blindly to manage what goes on in the physical realm.

For instance, a truck is scheduled by a generative AI tool to deliver a load onto newly laid concrete, which doesn’t have enough time to cure before the truck’s arrival. After the slab cracks, third-party property is damaged, and company operations halt, the logistics company receives a hefty lawsuit.

Circling back to Air Canada’s critical case, company leaders might think generative AI is only harmful when it’s customer-facing. However, cases like these demonstrate that even internal AI tools can carry outbound risks that lead to liabilities. A digital scheduling glitch translates to millions of dollars in structural damage that every company can avoid by vetting and training its tools with more rigor.

The Path Forward

Instead of scaring innovative companies away from generative AI, these common liabilities should simply encourage them to use it more critically. The new technology is here to stay. It offers a massive operational advantage, drawing a clear line between those stuck in the past and the leaders moving into the future. While the efficiency gains are far too big to ignore, adopting AI must be paired with education and sharp risk management strategies that foster responsible integration.

It’s also worth noting that standard corporate insurance policies—most of which were written years ago—often contain massive gaps when it comes to AI-driven errors, IP theft, or operational damage. In fact, many traditional General Liability, E&O, and Cyber renewals are actively adding strict AI exclusions.

That is why connecting with your insurance broker to audit your coverage is a vital step in protecting your company’s assets. To solve this exact problem, we’ve partnered with Testudo to offer a new, standalone Third-Party Generative AI Liability Insurance policy. Backed by Lloyd’s of London capacity, this policy is specifically built to fill those traditional coverage gaps. Whether a chatbot gives a customer costly misinformation (similar to the famous Air Canada incident), or an internal AI tool accidentally triggers an IP infringement or privacy leak, Testudo provides explicit cover for the financial and legal fallout.

Building an AI-forward company requires understanding where the guardrails need to be placed, training your team on proper tech practices, and securing the right safety net. We trust these five liabilities will help guide your governance efforts, and our team is here to help you secure the exact coverage you need to innovate with confidence.

Insurance Rebuilt, End-to-End

We've supported startups from stealth mode to huge funding rounds.

Get a Custom Quote