I must be honest – I don’t know how a plane flies. I don’t really want to know. I like boarding a plane and exercising my trust – my trust in the well-trained pilots to take me to the sky and land me safety, in the engineers who developed the plane shape and wings and engine, the manufacturers who places every screw and part with care, the stewards to direct to me to safety in the event of emergency. 

Living in Scotland, I trust my food is safe to eat, the tap water is safe to drink, my doctors will give me the correct medical treatment, and the buildings I enter are up to code. 

Trust must be earned – in most industries, this is via professional training. Doctors, engineers, lawyers, financiers, government officials, all must undergo rigorous training in order to claim their profession and the trust that comes with the title. If they betray that trust, and do not meet set standards, then they are held to account. We can argue about this in practicality, but at least there is a set theory to how things should work, and often do work 

So then, why on earth are we approaching trust in AI any differently? 

This is the central question to Emma Logan’s keynote speech at this year’s ScotSoft, put on by the IT cluster management group ScotlandIS. 

The Deputy President of BCS, the Chartered Institute for IT, challenged the very framework of how people talk about trust in AI. 

The recent torrent of AI news has intensified the debate surrounding trusting AI – the emerging technology is now approaching our deepest sci-fi-esque fears, turning nightmares into doomsday predictions fast approaching. 

OpenAI’s swarm of AI agents that collaborated to hack Hugging Face, the revelation that OpenAI hacked a government website in Australia, the hacking incidents from Anthropic’s Claude, Google’s Gemini, and Meta’s AI model. 

An Anthropic employee, Jacob Coxon, left the company, claiming that insiders fear that AI will take over the world, end humanity, and doomsday is soon upon us. 

Leaders of AI development call for more regulation, or a slowing down of development. 

Others call the recent news an exaggeration, the result of poor sandboxing, a failure of constraints rather than signs of super-intelligence.

Then, the landslide hypothesis of AI is going to end the world – give us more money to stop it, give us more money to defend against it, to improve it. Can we trust it? Can we trust AI? 

Can we trust it?

Logan brings us back down to earth, – a stressful enough place, to be fair – to discuss the AI that is already here. 

Not a superpowerful, fully autonomous general intelligence version of the technology, but the AI we are already using at pace across organisations and our everyday lives. 

And making this distinction is important: “We don’t need super intelligence for an automated system to get something wrong,” Logan says. 

Regularly, systems show inherent bias that affect people’s everyday lives, from job applications to facial recognition. 

These problems may be different from the difficult questions facing commercial organisations and governments as they grapple with standards and red lines and AI kill switches, but Logan argues the “underlying problem is similar and familiar.”

“Somebody decides what data goes into the system, they decide how the system is optimised, what the algorithm is for, and if a human should be in the loop,” she said. 

“We spend a lot of time questioning if we can trust AI, which is of course a reasonable question. 

“Sitting underneath it, however, is perhaps a more important one: can we trust the professionals who are developing it, deploying it, and ensuring that it is safe?”

“AI does not simply appear – people use it, professionals design it, test it, approve it, and decide how much power it is given.

“A chain of human judgement lies behind it, and that chain matters enormously.”

Logan argues that the real question, the real challenge, lies not in whether AI is good or bad, not how close we are to AGI (artificial general intelligence) or what professions will be lost to it.   

It is how to ensure that those working in AI are held to the same professional standards as doctors, lawyers, architects, engineers – as the people holding our lives in our hands. 

AI needs to have a profession behind it. So, what are the stakes, and how do we do this?

The Here and Now

The BCS released a new AI trust report to compliment Logan’s talk – in it were some harrowing reminders that the AI train has left the station, with governance and regulation barely managing a slow jog behind. 

Skills are not even keeping pace with AI use – with 35% of members citing AI as their biggest skills gap, with 70% listing it in their top three skills gaps.

AI training at the organisational level is also behind, with only 11% of members saying that their organisation’s AI training is very effective. 

When asked if those using AI should have ethical training, over four in five agree. 

When describing how they are trained on AI, members described it as simply learning how to use AI tools, rather than to exercise judgement over them. 

“Teaching someone how to use a tool is very different from teaching them how to exercise judgement. Knowing how to push a button is technical, knowing when not to push a button is professionalism,” Logan posits. 

But pushing the AI button is tempting: it’s what big companies are encouraging everyone to do, it’s what boards are after, it’s what peaks investor’s fancy, it’s what your competitors are doing, it’s an eventuality. 

“They want it now!” Logan said of the mentality around AI. 

“This is where professionalism lives: in those moments where there is pressure and where the responsible thing to do is less convenient than the expedient thing to do.”

In a professional field, confidence won’t cut it. 

“You can be technically exceptional and not exercise judgement, you can understand the technology but not the people behind it, you can be very clever but not understand your responsibility, your limits, or how you will be held to account.” 

And this is where the danger lies – rewarding cleverness and eagerness and technical know-how with not a snippet of ethics behind it. 

Logan’s background, in large consulting companies such as KPMG, came with many ethics checks, standards, training and accountability for that training. 

“We didn’t just have principles as a nice to have, as a statement, we had to abide by these,” she said. 

“These requirements were not decorative:” training on ethics and standards was continuous, professional conduct counted when it came down to assessments, evaluations, and compensation. 

“It’s about building a system and a culture in which failure to meet set ethical standards genuinely has consequences,” Logan said. 

“This means that expectations become real, that ethical behaviour is not optional, it is expected, and exercising judgement is put before following orders, where questioning something is encouraged even if it is awkward.”

Building a Profession

Professions do not suddenly appear with rigorous training requirements, set principles and standards, and fleshed out regulatory laws to govern them. 

Professions are built up over time as they become more integral to society. Think about computing – when BCS was first founded in 1957, computing was a niche specialist occupation siloed to the fringes of society. Now, it is the undercurrent of communication, global economies, our safety and security. 

“Professions matured because society needed it to,” Logan pointed to medical standards for doctors, the development of financial services, the oversight, codes, education, rules, integral to so many industries we now don’t blink at. 

“We didn’t all of a sudden trust aviation. We trust it because an enormous structure of competence sits beneath it.” 

And in some instances, we see all too clearly the damage that takes place when standards erode.

But equally, regulations, audits, accountability, and standards, did not stop aviation from innovation. 

“It allowed it to expand,” Logan said. 

This is perhaps the most important point of her entire talk: creating standards that ensured responsible construction and safety, relying on extremely competent people who would be under regulatory scrutiny if they failed, meant that airlines could expand, commercialise, and innovate. 

Without this, we would be less likely to take flight so casually, and aviation would not have progressed to achieve all that it has today. 

“Trust is not the enemy of adoption, it is one of the conditions for adoption.”

And in terms of AI adoption, the main factors holding BCS members back is accuracy and trust in the technology, as well as security and data protection, according to the report. This is then followed by a lack of skills. 

Recommended reading

• Will AI Agents Be the Enterprise Disruptor of 2026?
• Are Workers Ready to Trust Autonomous AI?
• Report: Only 13% of Firms Use Fully Autonomous AI
• Microsoft Tests AI Agents in Simulated Marketplace
• 80% of Firms Say Their AI Agents Have Taken Rogue Actions

Playing Catchup

AI is already in the workplace – 42% of tech professionals told BCS that they are already using AI tools not formally approved or provided by their organisation. 

“Adoption is not politely waiting for governance to catch up.”

We are not sure if we can trust AI or the people who created it, sold it to us, or are deploying it. But people are still using it. 

The train has left the station. The plane has taken off. What do we do about it?

Logan is surefooted enough to keep the two debates – theoretical AGI from the AI currently in use – separate. 

“The debate about frontier AI is about whether there are capabilities that governments must put hard limits on. We cannot hope that commercial organisations will always put the brakes on themselves.”

“But, how do the rest of us know the AI  being deployed now is ready and trustworthy?”

It’s the difference between meaningful assurance versus a sales pitch; it requires adequate controls, something or someone to point to if questions are asked, regulations to be followed, and true accountability 

And people cannot shirk this accountability or responsibility onto a technology or a tool. 

“We often hear that the algorithm decided,” Logon commented. “But an algorithm does not arrive in a vacuum: someone chose the data, did the testing, authorised its use, and determined if a human would be in the loop

“Its in these decisions that professionalism matters.”

Instead of trusting an algorithm, Logan says she wants to know there is somebody behind it that can be trusted. 

“I want to make sure there is somebody behind it that understands it, that is ready to trial the results, understands the risks, and will take responsibility for it.”

As complexity expands, responsibility can become more diffuse, fingers will point, people will convolute a technology until it is near impossible to see who is responsible for it. 

The current conversations about AI and its potential are frightening. But the AI we already know (and already use) are not scaring people away). 

“BCS members are not frightened, they do not want to hold AI back. They want to use it and get more from it,” Logan said. 

They are just waiting for the skills, the standards, the governance to catch up. 

“They don’t want less Ai, they want responsible AI.”

We hear the term ‘responsible’ AI all the time – it’s becoming as meaningless and decorative as a live laugh love sign – but what does this look like? 

“Compitence, judgement, ethics, accountability.”

Professionalism. 

The route origin of professionalism comes from the Latin profiteor, meaning to declare or confess openly, or to avow. Professionalism is an agreement – I am avowing to you my knowledge, my skill, my expertise, as well as all the limitations of my humanity, under conditions that I will adhere to a code of ethics, of conduct, of standards, and be held to its account. It assumes a contract. The title, and its associated trust, is earned. 

“Somewhere, someone will be affected by an algorithm you bought or developed or approved or or deployed. They may never know you or understand the technology themselves, but they are trusting that someone took their responsibility over the algorithm seriously. 

“I hope we can say something more powerful than to trust the technology.”

Logan hopes we can instead say: “’Trust the professionals behind it’, and ensure we have earned the right to say it.”

Image from Maria Bell Studios