Intervention

AI and ethics, how to fuel innovation while protecting people

AI adoption is growing rapidly, but without oversight and ethical principles it risks fuelling discrimination and mistrust among users

by Yari Franzini*

10 December 2025

Aggiungi Il Sole 24 Ore
ai preferiti su Google

3' min read

Translated by AI

Versione italiana

3' min read

Translated by AI

Versione italiana

According to IDC, European spending on artificial intelligence will reach USD 144.6 billion by 2028, with a compound annual growth rate of 30.3%. However, although 66% of people worldwide regularly use AI systems, only 46% fully trust them. In fact, without close oversight, the very tools designed to promote progress can distort results and exacerbate the presence of biased information, thus departing from the principles that ideally underpin them: fairness, transparency and accountability.

Responsibility as a starting point

Accountability is the fundamental pillar of an ethical implementation of artificial intelligence. For example, a bank could use AI to determine whether to approve or deny a loan application. If the applicant receives a 'poor credit history' notification and the bank denies the loan, the banking institution takes responsibility for the AI's decision to reach this conclusion.

When actions or decisions taken by AI are not communicated openly, as in the case of bank notification, trust and transparency between the parties involved is compromised. This is why it is vital to ensure that the human agents who create the AI lifecycle - designers, developers and implementers of the system - are held accountable for its proper implementation by clearly defining tasks and roles and providing each figure involved with the tools to justify the decisions supported by the algorithms.

Organisations are therefore called upon to develop a 'Responsibility-by-Design' approach to AI, aimed at implementing systems based on ethical design principles from the outset. Not only that, in order to comply with evolving standards, frequent evaluations of AI outputs are required, in an ongoing effort to avoid bias, abuse and unforeseen issues at the design stage.

Tackling prejudice and promoting equity

From facial recognition software that misidentifies particular demographics, to recruitment tools that discriminate on the basis of gender or ethnicity, AI systems have repeatedly demonstrated the urgent need for fairer and more transparent technologies.

Many organisations already make use of algorithms to provide initial triage, e.g. for screening candidates in recruitment processes. In this case, it becomes essential to monitor and review them regularly to identify and remove any discriminatory tendencies in the evaluation of resumes. Indeed, to ensure that automated decisions are informed by logic, ethics and empathy, it is imperative to maintain a 'human in the loop' approach, in which there is always a person who supervises and can intervene in the operations of the AI to ensure that the results are unbiased.

Trust grows when transparency is evident

Continuing the recruitment example, imagine applying for a job and discovering that your application has been evaluated and rejected by an AI system, without even reaching a human recruiter. This lack of transparency can lead candidates to lose confidence and believe that they did not have a fair chance in the recruitment process.

Organisations must be able to clearly show how AI systems evaluate applications and make decisions, providing full explanations of the standards and procedures implemented. It is therefore necessary to have easily accessible channels for user comments or appeals and to ensure that users know whether AI conclusions are final or subject to human review.

Equity and accountability in IA implementation are further enhanced by open internal governance frameworks, such as those that explicitly delineate the jurisdiction of IA ethics committees.

Inhancing privacy and security

AI can collect and analyse huge volumes of data at previously unimaginable speeds. This power, however, carries a high risk of privacy breaches, as demonstrated by McKinsey, which found that 8 out of 10 European citizens express concern about who can access their personal data.

Respect for individual rights goes beyond legal compliance and, in the development of AI, the protection of sensitive and personally identifiable data should be a top priority.

This can be achieved through AI literacy courses and by continuously updating the skills of both individuals and the entire organisation.

The adoption of ethical standards in AI is no longer deferrable

The advancement of AI technology often exceeds the capacity of existing regulations and ethical standards. Delaying action exposes companies to the risk of its harmful or unpredictable use, as AI now concretely affects many areas of daily life, such as healthcare, work, privacy and security.

Ethics and security have become strategic priorities for the EU, which aims to make significant progress towards the responsible development of AI models and to provide citizens with the necessary protection mechanisms. Proof of this is the Code of Conduct for General Purpose AI (GPAI) recently launched by the European Commission.

AI development is by nature distributed, involving a variety of institutions and countries. By establishing common moral principles, we can all take responsibility for preventing abuse and promoting useful applications that build global trust.

*Group Vice President Southern Europe of Cloudera