Digital Literacy Has Changed: Knowing When NOT to Trust Technology

Digital literacy is being reconsidered at a particularly important moment in history. From 8 to 11 September 2026, Unesco’s Digital Learning Week is bringing together discussions around the future of education and technology at UNESCO Headquarters in Paris as well as online.

2026’s theme, “Education in the age of AI: Facts | Frictions | Frontiers,” examines how artificial intelligence is reshaping education, including the rise of synthetic knowledge, public interest AI and digital sovereignty.

Although the conversation is mainly centred on education, it raises a question that extends well beyond the classroom, “What does it actually mean to be digitally literate when technology can increasingly generate information, make recommendations and influence decisions for us?”

For years, digital literacy has been about knowing how to intelligently navigate technology. It meant knowing how to use a spreadsheet, search for information, send an email, work with business software or find your way around a new platform.

As technology became more sophisticated, digital literacy expanded to include things like cybersecurity awareness, data privacy and the ability to distinguish credible information from misinformation.

But technology has changed at a breakneck speed. The systems people interact with today are increasingly capable of generating information, making recommendations, identifying patterns and, in some cases, even taking action on their behalf. All of this has changed what it means to be digitally literate.

It's no longer about just knowing how to use technology but more so about knowing when to trust the technology and, more importantly, when NOT to trust it.

Digital Literacy Used to Be About Doing

Traditional digital literacy used to be largely about being competent. In other words:

• Could you operate the technology effectively?
• Could you find the information you needed?
• Could you troubleshoot a basic technology-related problem?
• Could you know the difference between a legitimate website and a suspicious one?

The assumption was relatively straightforward: technology was a tool, and the human was the decision-maker. But, the relationship is becoming much less clear because of the advent of AI.

AI systems can technically operate almost autonomously. These systems can draft reports, summarise documents, analyse data, recommend actions and generate answers that sound authoritative (even when they are incorrect).

Automation can make decisions without a person reviewing every individual step. Enterprise software can increasingly predict what a user needs before they even ask for it. The technology is no longer waiting for instructions but increasingly participating in the decision-making process. That means digital literacy needs another layer: judgement.

Ward Against Being Confidently Incorrect

One of the biggest challenges with modern technology is to remember that confidence and correctness aren’t the same thing. Like they say, one shouldn’t be loud AND wrong.

A system may be able to produce a quick answer, present it in a polished way and attach a convincing explanation about how it came to that conclusion without that answer even being accurate in the first place. This is particularly important when it comes to generative AI.

People look towards things like fluency, confidence and familiarity to judge whether information can be trusted. But AI systems don't necessarily work according to those same signals which presents a conflict.

A response can sound certain without being certain. A recommendation can be plausible without being appropriate. A summary can be well written while leaving out the information that matters most. And lastly, an automated decision can be technically valid but be wrong for the particular context in which it's meant to be applied.

The World Economic Forum has argued that AI literacy increasingly requires people to understand when NOT to trust AI, rather than simply knowing how to interact with it. That distinction is becoming one of the defining digital skills to look out for in the upcoming few years.

The New Digital Skill: Knowing When to Stop

Consider an employee using an AI assistant to summarise a lengthy contract. The system produces a concise summary in mere seconds. A user who you might see as digitally competent in all other areas might accept it at face value because the output looks professionally done.

However, a digitally literate user in the modern sense of the phrase would look at it from a more skeptical angle and ask questions like: 

• What information might have been omitted?
• Does the system have access to the entire document?
• Is the interpretation consistent with the original wording?
• Does this final product carry legal or financial consequences?
• Should a subject-matter expert review the end result?
• What happens if the AI is wrong?

The difference is subtle, but still very important. The first person knows how to use the technology. The second understands usage alongside the boundaries of the technology. And that is a much more valuable skill in today’s age.

Not Every Decision Deserves the Same Level of Trust

This does not mean organisations should be scared of using AI or automation. It simply means employees on all levels need to become better at understanding where utter confidence is appropriate to apply. For example:

• An AI-generated list of content ideas for a marketing campaign is relatively low risk and easy to sift through before acting upon.
• An AI-generated summary used to brief an executive on a topic might require closer inspection.
• An automated recommendation affecting a customer's credit, employment, healthcare or access to services requires considerably more.

The technology may be identical but the appropriate level of trust is not. This is why digital literacy now increasingly intersects with risk management.

NIST's AI Risk Management Framework, for example, treats trustworthy AI as involving characteristics such as validity and reliability, safety, security, accountability, transparency, explainability, privacy and fairness.

It also emphasises the importance of defining and documenting human oversight. In other words, trusting technology should not be a twofold decision that is wrapped up in context.

The Rise of Verification Literacy

There is another skill emerging alongside digital literacy, namely verification literacy. Someone with verification literacy is able to evaluate online information with a critical eye, spot false claims, and confirm whether a source is true and accurate.

This concept is particularly relevant in an environment where synthetic content can look increasingly authentic (think AI deepfakes). For years, people were taught to ask whether information was credible. Now they also need to ask how that information was produced.

• Was it written by a person?
• Generated by AI?
• Pulled automatically from another system?
• Based on current data or an outdated dataset?
• Verified by an expert?
• Produced from a source that the user can actually inspect?

But even these questions introduce another problem: people are not necessarily very good at identifying whether something is genuinely AI-generated. AI detectors can produce false positives, while people can mistake authentic content for synthetic knowledge simply because it appears unusual, polished or unfamiliar.

A 2026 study of human perception of audio deepfakes illustrates the problem. Researchers found that people's ability to identify fake audio had changed very little when compared to a 2021 baseline, but their accuracy at identifying genuine audio fell from 72.7 per cent to 64.1 per cent. In other words, people were not simply becoming better or worse at spotting AI; they were becoming more likely to distrust content that was actually real.

That matters because verification literacy should not become another form of pattern recognition. An output looking or sounding like AI is not, by itself, evidence that it is AI-generated. Nor does content appearing human make it trustworthy. The stronger approach is to examine the provenance, source, evidence and context behind the information.

The challenge becomes even greater as organisations increasingly work with synthetic knowledge. Instead of information that has been received from directly observing or recording the real world as is, synthetic knowledge is information that has been created by combining, generating, or transforming existing information.

The World Economic Forum has described verification literacy as understanding how information becomes trustworthy, rather than judging truth simply by what it looks like at first glance. That is an important distinction.

The new digital literacy question has therefore become whether you can establish whether information deserves your confidence rather than simply identify where you found it or what it appears to be.

Automation Exposes Another Literacy Problem

Automation is the other challenge of digital literacy alongside AI. Enterprise environments are increasingly built from interconnected systems, APIs, automated workflows, cloud platforms and data pipelines.

An employee might trigger a process without ever seeing everything that happens afterwards. A customer record is updated in one system. That update triggers another workflow. The workflow calls another service. That service applies a model or business rule. The resulting decision is passed somewhere else.

By the end of this, the person who initiated the process may have little visibility into how the final outcome was produced. This creates a different kind of digital literacy problem.

Employees don't necessarily need to understand every technical component. But they increasingly need to understand where their visibility ends. Knowing what you cannot see can be just as important as knowing what you can.

The Danger of Automation Bias

A psychological problem also rears its head. When technology consistently produces useful results, people can become less inclined to challenge it. On a long enough timeline, the confidence in the results are compounded.

This is sometimes described as automation bias. This is the tendency to favour automated recommendations or decisions over one's own innate judgement, something that was there way before AI and automation came along. The danger of this is obvious.

The more reliable a system appears, the less likely people may be to intervene when it makes an unusual recommendation that doesn’t quite feel right. It then creates a paradox that the better technology becomes, the more important human judgement can and should become.

This is not because humans should manually repeat everything technology does for verification, but because humans need to recognise which situations don’t require technology's assumptions.

NIST's AI guidance similarly stresses that human oversight needs to be structurally defined rather than treated as an automatic safety mechanism. A human sitting somewhere in the workflow is not necessarily meaningful oversight. That human needs the authority, information and confidence to challenge the system.

So, What Should Modern Digital Literacy Look Like?

If digital literacy is changing, organisations need to change how they teach it. The goal is no longer simply to make employees proficient with new tools, but to develop the judgement needed to use those tools responsibly.

Instead of focusing exclusively on software proficiency, training should include questions such as:

1. What is this system actually doing?

Users should understand the basic purpose and limitations of the technology they are using. They don't necessarily need to become AI engineers. But they should understand whether a system is generating, predicting, retrieving, recommending or automatically acting.

2. Where can the system be wrong?

Every technology has failure modes (essentially everything can fail). Users need to know what those failure modes look like and what the warning signs are to prevent or prepare to fix it.

3. What information is the system using?

The quality of an output depends partly on the information behind it. What are the sources being used? Users should understand whether a system is working from current, complete and appropriate data.

4. When should I verify the result?

Ideally, every time you get an AI response or output, you need to scrutinise it. But not every AI response requires a lengthy deep dive.

Some decisions require more verification than others. Digital literacy should help employees learn how to naturally distinguish between minimal and detailed review.

5. When should a human take over?

There should be clearly understood circumstances where automation stops and human judgement and oversight begins.

6. What happens if the system is wrong?

This may be the most important question of all. If a mistake has minimal consequences, automation may be appropriate and valid.

If the mistake could create financial, legal, security, safety or reputational consequences, there should be a lower threshold for when human review steps in. In other words, it should be easier for the trigger to be pulled.

How to Test Your Employees’ Digital Literacy

If digital literacy is increasingly about judgement rather than simply technical proficiency, organisations need to rethink how they assess it.

A traditional digital literacy test might ask employees whether they know how to use a particular application or complete a particular task. Those skills do still matter, but they don’t necessarily reveal if someone can recognise when technology should be questioned.

Instead, organisations could test employees with realistic scenarios. Some digital literacy testing options to try include:

• Give them an AI-generated answer containing subtle inaccuracies. Can they identify what needs to be checked?
• Present two conflicting outputs from different systems. Do they know how to investigate the discrepancy?
• Ask them to use AI to summarise a complex document. Do they verify the summary against the original?
• Give them an automated recommendation with incomplete data. Do they recognise that the recommendation may be unreliable?
• Present a high-stakes decision alongside a low-risk decision. Do they understand that the two require different levels of human oversight?
• Ask them to explain why they trust a particular output. Can they identify the source, assumptions and limitations behind it?

That final test may be the most revealing. An employee who can explain why they trust an output has demonstrated more than an ability to reach the right answer. They have demonstrated that they have stopped to evaluate it. This is increasingly important as AI can make information easier to consume while potentially encouraging people to accept outputs without critically examining them.

The objective isn't to catch employees out. It’s to establish whether they understand when technology can be used confidently, when its output needs verification and when a human needs to take over.

This could also provide organisations with a more useful measure of AI readiness than simply asking employees whether they feel comfortable using AI tools.

The Goal Isn’t Distrusting Technology

There is an important distinction between healthy scepticism and overall distrusting technology. The objective is not to teach employees that AI cannot be trusted. It's to teach them that trust should be earned and measured.

Technology is extremely useful because it can process more information, identify patterns and perform repetitive tasks at a scale that humans can’t. But the mistake is assuming that that capability equals reliability in every situation.

Of course, a calculator is excellent at arithmetic. That does not mean it knows whether the equation you entered in the first place (the input) was the right one. It simply provided the best information based on what was given to it.

An AI system can produce an excellent summary. That does not mean it includes everything of importance for you to make a high-stakes decision (unless it has been given a goal to work towards).

A predictive model can identify a strong pattern. That does not mean the pattern will remain true when circumstances change.

All of this has resulted in the human role shifting. It's moving away from simply using the technology towards judging what the technology produces.

The Most Digitally Literate Employee May Be the One Who Says “I’m Not Sure”

For organisations, this could require a significant internal cultural shift. Traditionally, employees have been rewarded for being efficient, decisive and technologically capable. But an AI-enabled workplace may also need to reward people for knowing when to take a beat.

In this type of environment and culture, you need:

• The employee who asks, “Where did this figure come from?”
• The manager who says, “Let's check that recommendation before acting on it.”
• The analyst who notices that an automated result does not match what they know about the business.
• The security professional who questions an apparently legitimate system behaviour.

Don’t look at these as signs of employees being resistant to technology. They are actually signs of digital maturity. As AI becomes more embedded in enterprise systems, the ability to challenge technology may become just as valuable as the ability to use it.

Digital Literacy Is Becoming a Judgement Skill

The first era of digital literacy was about access. Then it became about competence. Now it's increasingly about discernment and good judgement. Knowing how to use a technology is no longer enough when that technology can independently generate information, influence decisions and take action.

People need to fully understand technology’s capabilities as well as understand its limitations. They need to recognise when its output may be plausible but still uncertain enough to cast doubt. And they need to know when human expertise should take over.

In an enterprise environment it means digital literacy can no longer be treated as a one-time training exercise. As AI capabilities evolve, employees will need to continuously develop their ability to evaluate, question and work alongside these systems.

That does not make technology less valuable. It instead makes the relationship between humans and technology more sophisticated.

The future of digital literacy may therefore have less to do with knowing every button, feature or platform and more to do with knowing when to trust the machine, when to question it and when to take the decision back into human hands.

Because in a workplace increasingly filled with systems that can answer almost any question, the most valuable digital skill might simply be knowing which answers deserve to be believed. Getting a fast answer isn’t necessarily always the mark of superiority.

As AI continues to reshape how organisations work, developing that judgement will become just as important as learning how to use the technology. If you want to know how to up your enterprise’s digital literacy even more, follow EM360Tech’s AI feed to keep up with the latest developments changing how enterprises use, govern and trust AI.