Are We Adopting Artificial Intelligence Faster Than We Are Discussing Its Consequences?
- Published
It was hard to follow the news about Artificial Intelligence this week without running into some discussion of its risks.
Dario Amodei, CEO of Anthropic, argued that the development of the most advanced models needs to slow down to give safety measures time to catch up. Sam Altman of OpenAI, Elon Musk of xAI and other industry leaders voiced support for the need to reduce the pace, at a moment when researchers and industry figures once again put scenarios of loss of control and potentially catastrophic risks on the table.
Some consider these risks serious enough to justify immediate intervention, and some see a number of these scenarios as speculative or exaggerated. The industry itself is far from having a consensus on how to respond to the problem.
I do not know whether Artificial Intelligence will ever pose an existential threat to humanity. Right now, nobody knows.
But there are far less speculative consequences that we can already discuss.
Every day we put our organisations' documents, data, code and information into Artificial Intelligence systems. We are starting to use them to automate tasks, analyse information and support decisions. And, increasingly, these systems no longer just answer questions: they start to take action.
Are we developing our ability to use this technology at the same pace at which we are increasing its capability?
This question became even more interesting to me because it cuts directly across a piece of research I carried out with Professor Rui Pedro Marques.
In the study Research Trends in Artificial Intelligence in Education and the Ethics of Privacy, recently published by IEEE in the proceedings of ADACIS 2025, we analysed ten years of scientific research on Artificial Intelligence, privacy and education.
And what we found helps to show that this discussion did not start this week.
The growth happened very quickly
In the study we analysed the scientific research published between 2015 and 2024 on Artificial Intelligence, privacy and education.
And one of the results is particularly clear.
Until 2020, fewer than 60 papers a year were published on these topics. From 2021 onwards the trend changes quickly and, in 2024, we find more than 700 publications, more than double those recorded in 2023.
This does not mean, of course, that there were no concerns about AI and privacy before. It means that the recent expansion of Artificial Intelligence, and of generative AI in particular, has placed these questions at the centre of the discussion.
The very vocabulary of the research has changed. Among the most frequent terms we now find ChatGPT, Generative AI, Ethics, Privacy and Data Privacy.
A discussion that does not belong to education alone
Our research was set in the context of education, but many of the questions we found are today exactly the same ones I raise when I work with companies.
What information can we put into an AI tool?
What happens when we upload an internal document?
Can we use customer data?
Who validates an answer produced by AI?
And who is responsible when a decision supported by these systems is wrong?
Privacy, transparency, consent, bias and accountability appear again and again in the literature we analysed.
These questions are less and less theoretical.
The more AI stops being just a chat window and starts working with our documents, data and processes, the more important it becomes to understand how we are using it.
Adopting AI also means creating criteria for using it
I have argued, in the companies I work with, that experimenting with these tools matters. Creating fear around the technology does not solve the problem.
But it also does not seem enough to me to make a tool available and assume that everyone will automatically know what they can and cannot do with it.
We need to create criteria, define internal rules, train people and keep human oversight over what really matters.
The technology will keep evolving quickly. It will probably be hard for organisations to keep up with every tool, model and feature that comes along.
That is why an important part of preparing for this change is not just about learning to use the next tool.
It is about developing, in people and in organisations, the ability to understand when to use AI, with what information, for what purpose and with what responsibility.
And that is not something you learn from a 60-second reel by an AI influencer or guru.
The study "Research Trends in Artificial Intelligence in Education and the Ethics of Privacy", by André Bacêlo e Costa and Rui Pedro Marques, was presented at the 2025 IEEE International Conference on Advances in Data-Driven Analytics And Intelligent Systems (ADACIS) and is published in IEEE Xplore.