We use cookies to personalise the website and offer you the greatest added value. They are, among other purposes, used to analyse visitor usage in order to improve the website for you. By using this website, you agree to their use. Further information can be found in our data privacy statement.



Recruiting without recruiters: opportunities, risks and regulatory obligations

​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​published on 20 May 2026 | reading time approx. 7 minutes


The job interview, as we have known over the last few decades, is undergoing a radical transformation.
Amazon has in fact launched ‘Connect Talent’, a system that conducts job interviews completely autonomously, 24 hours a day, seven days a week. It is an innovation that promises efficiency, but which raises legal issues, risks of discrimination and employment concerns that companies cannot afford to ignore.

The Amazon announcement​

Amazon has recently unveiled two new AI-integrated software solutions for human resources: ‘Connect Talent’ and ‘Connect Decisions’. The first is designed to manage the entire recruitment cycle autonomously – from identifying candidates and conducting initial screening to carrying out interviews through voice interaction – without any direct human intervention. The second supports internal organizational decision-making, particularly in relation to supply chain management. 

Both tools were developed within the framework of so-called ‘humorphism’, a term coined by Amazon itself. According to the company –  which claims to place human beings at the center of technological innovation –  the concept refers to an artificial intelligence capable of learning the language, rhythms, and dynamics of human work, rather than forcing people to adapt to the machine.
 
However, whilst it is true that people do not have to adapt to the machine, it is equally true that the machine can replace people. This second premise appears to be confirmed by Amazon’s recent history, as the company has cut around 30,000 corporate roles in recent months, explicitly linking some of the cuts to the savings achieved through automation. 

The European regulatory framework: clear and non-derogable obligations​​

The adoption of systems such as Connect Talent, which is the focus of this discussion, is not without legal constraints, particularly within the European Union. 

Regulation (EU) 2024/1689 on artificial intelligence (‘AI Act’) classifies AI systems intended for use in the recruitment or selection of individuals as high-risk systems, including those used to analyze and filter applications and assess candidates. This classification imposes specific and non-derogable obligations for the users (deployers) of such systems, including: documenting the functioning of the algorithm and the selection criteria applied; ensuring technical and organizational compliance prior to deployment; ensuring meaningful human oversight of final decisions; clearly informing candidates about the use of AI in the selection process; and carrying out an assessment of the processing of personal data in accordance with Article 35 of Regulation (EU) 2016/679 (‘GDPR’).

These obligations are accompanied by a restriction that has been in force for several years: Article 22 of the GDPR prohibits decisions based solely on automated processing that produce significant legal effects on natural persons – except in limited circumstances subject to specific safeguards, including the candidate’s right to obtain human intervention, express their point of view, and contest the decision. 

In the context of recruitment and selection activities, therefore, an automated rejection without human review is, in principle, already prohibited under current law. Connect Talent, which claims to be able to manage the entire recruitment process without human intervention, is structurally situated in this risk zone, unless the companies adopting it implement specific safeguards that the supplier alone cannot guarantee.

The Risk of Discrimination: A Problem Already Seen Before​​

One of the most critical aspects of AI applied to recruitment concerns the risk of systematic discrimination. Artificial intelligence models learn from historical data: if a company’s past hiring practices reflected disparities in gender, age, geographical or ethnic background, the system will tend to replicate and amplify those biases, not correct them. This is not a theoretical risk and, significantly, it is not even a new risk for Amazon.

The company itself, in fact, was forced to scrap its own automated recruitment system, developed in-house from 2014 onwards, because it had learned to systematically penalize female applicants. The system had analyzed CVs received over a decade – CVs that reflected the historical male dominance in the tech sector – and had translated that disparity into a selection criterion. Words such as ‘women’ in CV headings, or references to women’s colleges, automatically lowered the candidate’s score. The case was reported by the Reuters news agency in 2018, and Amazon later confirmed that the project had been discontinued.

What makes this episode particularly relevant today is not the technology itself, which has evolved considerably in the meantime, but rather the structural dynamic that produced the problem: a system trained on biased data reproduces those distortions, regardless of the designer’s intentions. Connect Talent, when applied to corporate contexts with unbalanced hiring histories, is exposed to the same risk.

In Italy, Legislative Decree 198/2006 (‘Equal Opportunities Code’) and Legislative Decree 216/2003 prohibit discrimination – including indirect discrimination – in access to employment. A discriminatory effect produced by an algorithm does not require intentional conduct: it is sufficient that the system generates unfavorable treatment toward a protected group. 

Particularly significant in this context is the allocation of the burden of proof. 

Under Italian anti-discrimination law, the ordinary evidentiary standard is effectively mitigated: it is sufficient for the candidate to provide factual elements capable of giving rise to a presumption of discrimination in order to shift the burden onto the employer, which must then demonstrate that the unequal treatment did not result from discriminatory criteria. In a process governed by an opaque algorithm, this burden can prove very difficult to meet – especially in the absence of documentation on how the system operates and on the outcomes of the selection process.

HR professionals and the unresolved issue of reskilling and upskilling​​

Alongside the regulatory dimension, there is also an employment-related issue that deserves careful attention. Connect Talent directly affects the professionals who manage the recruitment process, effectively removing them from the equation, as clearly emerges from Amazon’s own presentation of the system. Its stated target consists of companies handling large-scale hiring processes, where the volume of applications makes it difficult to manage every interview through traditional human-led methods.

The real issue is not whether automation will happen (it certainly will), but whether companies will choose to invest in retraining HR staff for different roles, including higher-level ones – such as supervising algorithmic outputs, critically analyzing the system’s outputs, managing exceptions and disputes, and ensuring regulatory compliance – or whether they will simply seize the opportunity to reduce staff numbers without offering alternative career paths. The second option is the most cost-effective in the short term. But it is also the one that exposes companies to the greatest legal risk: an AI system without qualified human oversight is, by definition, a system outside regulatory control, falling entirely within the scope of liability of the company that implemented it. 

From this perspective, investing in training for companies is not only an ethical choice, but also a concrete risk management measure, explicitly required by the AI Act.

What companies must do​

The adoption of tools such as Connect Talent should not necessarily be considered strictly prohibited from a legal standpoint. The point, in fact, is not the automation of the recruitment phase itself, but how it is governed. In practical terms, companies considering these tools will be required to:
  • clearly define the decision-making scope of the system and the escalation criteria triggering human review; 
  • thoroughly assess the quality of training datasets in order to identify historical biases, and implement periodic audits of system outputs to detect the emergence of discriminatory patterns during use;
  • carry out a data protection impact assessment regarding the processing of personal data;
  • establish procedures capable of providing meaningful explanations for the system’s evaluations and decisions; 
  • update contracts with suppliers to include explicit AI Act compliance obligations, particularly concerning technical documentation, risk management systems, and post-deployment monitoring; 
  • comply with the transparency obligations set out in both the European and local regulatory frameworks; 
  • implement structured upskilling and reskilling programs for the HR professionals involved in the process.

A recruitment process that entirely excludes human evaluation is currently incompatible with the European regulatory framework. A company that adopts these tools without proper oversight, without transparency, and without personnel retraining policies is likely to face exposure on multiple fronts simultaneously: regulatory, reputational, and legal – not to mention the ethical implications. 

Amazon’s ‘humorphism’ is a compelling narrative. But companies adopting these tools cannot base their strategic decisions on words and promises alone: they remain accountable to a specific regulatory framework, to a court, and – in the event of discrimination – to a burden of proof that the algorithm cannot fulfil on their behalf. 

AI governance in recruitment is not optional: it is already a legal obligation.​

from our newsletter

author

Contact Person Picture

Irene Pudda

Attorney at Law (Italy)

Associate Partner

+39 02 6328 841

Send inquiry

Profile

Contact Person Picture

Carlotta Caminati

Attorney at Law (Italy)

Associate

+39 02 6328 841

Send inquiry

Profile

Skip Ribbon Commands
Skip to main content
Deutschland Weltweit Search Menu