Responsible AI and Corporate Reputation: Can One Algorithm Destroy Years of Trust?

In February, a tribunal in British Columbia ruled that Air Canada had to compensate a customer who had been given wrong information about bereavement fares by the airline’s website chatbot. Air Canada argued, in part, that the chatbot was responsible for its own statements. The tribunal disagreed: the company is responsible for all information on its website, whether it comes from a static page or a chatbot. The amount at stake was small. The precedent and the headlines were not.
In January, a customer frustrated with DPD’s parcel delivery chatbot persuaded it to swear and to write a poem about how useless the company was. The screenshots went viral. DPD disabled the AI element.
Later in February, Google paused the image generation function of its Gemini model after it produced historically inaccurate images in an apparent attempt to ensure diversity. The criticism was intense, Alphabet’s share price fell noticeably in the following days, and the company’s CEO told employees the responses were unacceptable.
A new kind of reputational exposure
These incidents have something in common. None involved a catastrophic system failure. Each involved an AI system interacting directly with the public and producing something embarrassing, misleading or offensive. And each became a story about the company, not about the technology.
That is the new reputational dynamic. When AI faces customers, every output is a potential public statement by the company. Unlike a press release, nobody approved it in advance. And unlike a traditional software bug, the failure is easy for anyone to understand, screenshot and share.
Why the impact goes beyond the incident
The damage from such episodes rarely stays with the individual case.
Customers start to question the company’s other products. If the chatbot gives wrong answers, what about the rest?
Regulators take an interest, particularly where consumer protection or discrimination are involved.
Employees ask whether the company’s stated values are real.
Investors wonder whether management understands the risks of the technology it is deploying.
The Air Canada case also closes an argument many companies might have been tempted to make. An organization cannot distance itself from its AI. Legally and reputationally, the system speaks for the company.
Treating AI risk like other enterprise risks
The practical conclusion is that AI risk should be evaluated with the same seriousness and methods as financial, cyber and reputational risk. That means scenario analysis before launch: what is the worst plausible output this system could produce, and what would happen if it did? It means red teaming customer facing systems with people who try to make them fail. It means monitoring in production and a clear process to switch a system off quickly. And it means crisis plans that anticipate an AI incident before it happens.
Most of these practices exist already in other risk domains. The challenge is applying them to systems whose failures are less predictable and far more visible.
Leadership question: Should companies evaluate AI risk in the same way they evaluate financial, cybersecurity and reputational risk?


