Way back before AI was dominating the headlines and social media feeds every day, there was a story that went global partly because of the scandal it covered, and partly, I think, because of its tabloid-ready nickname, ‘Dieselgate’.

Firstly, it’s important we take a lesson from this and stop adding the suffix ‘gate’ to every scandal broken by the media. It worked for Watergate, it was novel and clever, it even worked a few times after that, but just like the Marvel movies, it’s time we started leaning into a little originality again!

Outside its name, the story hit the headlines because of one decision a group of company managers made. They intentionally programmed some of the brand’s vehicles to activate emissions controls only during lab testing so they could meet US and European emissions requirements.

US authorities discovered the Volkswagen stitch up. 

VW stock fell by up to 20% within two days, and a further 17% shortly after. The global CEO resigned. The Emissions Compliance Officer was charged with a criminal offence, as were six executives. The CEO was then also charged with criminal offences. The US CEO resigned. Civil suits were filed against individuals and the company. 

The scandal cost more than $30 billion, not including sharemarket losses.

In court in Germany, the CEO denied any wrongdoing, claiming he wasn’t aware of the decision.

But as CEO, the buck stops there.

The question is: if AI, rather than a group of human beings, had made that decision, would the CEO or the AI platform be held to account?

Canada Airlines seems to think AI should be the culprit.

Is AI more than an employee?

Back down memory lane again, this time, it’s 2022, and Jake Moffatt has reached out to Air Canada for support, only to find himself talking to an AI chatbot… seemingly a very empathetic one. 

Jake had lost his grandfather and needed to fly to his funeral. The chatbot assured Jake the airline had him covered, that he could purchase the full fare ticket and then get a discount afterwards. It said it was their bereavement policy and one of the ways they support and care for their customers.

Only it wasn’t true.

The chatbot unfortunately got the details mixed up and when Jake went to apply for his bereavement reimbursement as instructed, he was told no, in fact, he needed to apply beforehand. Denied.

The most interesting part of this case isn’t an airline that actually does have a potentially beautiful (if difficult to access) policy for those grieving, it’s what Air Canada’s defence was when Jake took them to court.

‘The AI is responsible for itself’.

While laws and precedents are different all around the world, it’s likely the airline would never have dared give the same argument should a poorly trained human administrator have answered Jake’s question incorrectly. Why? Because there is example after example of companies being held to account for the actions of their employees, regardless of management’s involvement in the situation. 

The VW incident is the perfect example. While the CEO’s strong denial of any wrongdoing was questionable, ultimately the company was held accountable for the actions of its employees. And as the CEO, he was personally held accountable for the decisions his teams made under his watch. 

That being the case, why would we treat early generation customer AI as more intelligent and accountable than a human? 

We wouldn’t.

Probably needless to say, Air Canada was ruled against. The airline is in fact responsible and accountable for the information provided by its own AI chatbot, on its own website, trained with its own information. 

Makes sense.

So, in a world where you, as leaders, are now not only accountable for the actions of your human team, from the most junior administrator to the most senior manager, as well as these relatively early iterations of customer-facing AI, where does that leave you? 

Leadership in the new age of delegation

It’s easy to look at a case like Air Canada’s and read their stance in absolute awe, until you realise how close it hits to home. Because at some point, every leader will face a similar question: if something goes wrong, who’s really accountable for a decision from a trained system rather than a trained human?

The answer is complicated, but the principle isn’t. The same way a leader is accountable for their people, they’re accountable for the technology they deploy, especially when it acts on behalf of their brand. 

In other words, AI may be intelligent, but it’s not autonomous in the legal or ethical sense. It doesn’t have intent (as it keeps telling me each time it hallucinates), it doesn’t have duty of care, and it certainly doesn’t have a defence lawyer. But you do. The accountability still lies squarely with the humans who choose to use it, the same as any other tool. 

And yet, that accountability is no longer as simple as it once was. When a team member makes a mistake, there is established HR, compliance, and governance processes to deal with it. There is a fairly surefire way to make sure it doesn’t happen again.  

But when a model pulls information from a training set no one has personally reviewed, and that information misleads or harms a customer, where does that fit into the risk framework?

It’s a new frontier for leaders who are now delegating both to humans and AI, and a new challenge for governance; one that requires leaders to move more actively beyond operational management and into ethical foresight and process oversight.

Redefining accountability before it redefines you

There’s a quiet irony in all this: the more AI is designed to take things off our plates, the more weight lands on leaders’ shoulders. 

But leadership has always been about navigating complexity, not avoiding it. The great leaders of this emerging AI era won’t be those who know how to code or who can explain what a large language model is, they’ll be the ones who understand how accountability needs to evolve and the integral part they have to play in that process.

That means thinking differently about governance. 

Historically, governance frameworks were built for compliance, designed to prevent misconduct, not miscalculations by machines. 

They only needed to move slowly and in reality, much of their guidelines could be fairly ‘set and forget’. They would remain accurate and up to date with quite infrequent review. 

Now, governance is about foresight and fluency: knowing what your systems can do, what they should do, and what lines they must never cross. It’s the same instinct that made a CEO question whether a marketing claim was too risky or a financial statement too rosy. Only now, the same discipline needs to be applied to models and algorithms, because those models are speaking on your behalf.

The best leaders are already adapting. They’re creating AI oversight committees, training staff to recognise when AI outputs should be questioned, and embedding risk assessments into every new deployment. 

They’re building transparency into process, documenting what the AI does, where it gets its data, and who’s responsible for reviewing its work. 

As a leader, your first job is not to ignore what’s happening, not to stick your head in the sand. 

It’s easy to think this doesn’t affect you yet because your business doesn’t use AI, but that’s likely the first thing you have wrong.

New research by TechRadar says three in five employees are using shadow AI; that’s AI tools not approved by your business and not written into your agreed processes. 

Whether you know it or like it or not, you are already accountable, and without comprehensive processes and governance, your business, brand and even your job are already at risk.

Time to take this bull by the horns!

From risk management to reputational resilience

If there’s one thing the Volkswagen and Air Canada cases make obvious, it’s that trust, once broken, is expensive to rebuild. 

Consumers today don’t distinguish between a bad decision made by an intern, a chatbot, or an executive. They see one brand, one company, and one failure. And they expect the leaders to own it.

This is why governance isn’t just about compliance; it’s about brand protection as well. AI governance frameworks are fast becoming as essential as financial audits, not because regulators and insurance companies necessarily demand them (though they soon will), but because public trust does. 

When an AI makes an error that harms a customer in any way, the reputational fallout can far outweigh the operational one, so it is essential guidelines are clear and practices vetted. 

And while the law, as is often the case, has fallen behind the tech, it is starting to claw back some ground and find its footing. 

As an example, the EU’s pending AI Act does away with claims like that of Air Canada and explicitly places legal accountability on the provider or deployer of AI systems.

It doesn’t matter whether the system was built in-house or licensed, if it impacts people, place or profit, the organisation behind it is responsible. 

It’s the same story in the United States, where the Federal Trade Commission has warned companies that AI claims and AI outcomes will be treated as their own, not their vendor’s. You can bet Australia is on the same path, and globally, most countries will follow. They have to. 

So, as the world adjusts to make space for AI, the risk for companies increases. The risk of failing compliance or breaking the law, the risk of harming someone or something, the risk of damaging the reputation of the business in a way that is irreparable. 

Are you ready to take on that risk?

Leading with foresight, not fear

None of this should terrify leaders… unless you plan to just ignore it.

If you have open eyes and minds, and you see it for what it is, it should energise you. Because AI, when paired with the right guardrails, can make processes and even some decision-making faster, fairer, and more informed than ever before. But the only way to reach that potential is through leadership that treats governance as a muscle, not a muzzle.

When we talk about AI governance (we know, not always the sexiest of words or concepts), we aren’t talking about a book of rules so tight people can barely move. We’re talking about guidelines that provide protection to anyone working with it or being served by it, and we’re talking about adaptive practices, because the guidelines you set today will likely be outdated within a handful of months at most, if you’re lucky.

The role of the modern executive isn’t to master AI, you have specialists and suppliers for that, it’s to master the conditions around it. To create cultures where people know when to trust the tools, when to question them, and when to take the wheel back. 

In times of change, evolution, uncertainty, this is where true leadership really shines. Not in panic, but in a clear purpose, knowing that trust, accountability, and human oversight are still crucial to holding everything together.

AI might be writing scripts, handling chats, and generating decisions, but it’s leaders and their teams who will be remembered for the integrity of the outcomes, one way or another.

It doesn’t matter right now if AI has access to more data than us, is smarter than us, is faster than us, the bottom line is that we, as humans, and you, as leaders, are still accountable for both the wins and the fallout it produces. The buck stops with us. 

Sources worth a read:

Volkswagen: the scandal
VW rocked by emissions scandal 
Volkswagen’s emissions scandal
Air Canada chatbot: what travellers should know
From emissions cheater to climate leader
Many workers using unapproved AI at work