The Rome Call for AI Ethics, Explained

The basic facts

Full title: Rome Call for AI Ethics.

Released: February 28, 2020 at the Pontifical Academy for Life, Vatican City.

Original signatories:

• Archbishop Vincenzo Paglia, then-President of the Pontifical Academy for Life
• Brad Smith, then-President of Microsoft
• John Kelly III, then-Executive Vice President of IBM
• Qu Dongyu, Director-General of the UN Food and Agriculture Organization (FAO)
• Paola Pisano, Italian Minister for Innovation

Type of document: A voluntary multi-stakeholder ethics commitment. The Rome Call is not a doctrinal document, not a treaty, and not a regulation. It is a declaration of shared principles signed by parties who agree to apply those principles in their work.

Length: Roughly 1,000 words, divided into a preamble, three "impact areas" (ethics, education, rights), and the six principles. It is short by design.

Approving body: Pope Francis received the signatories and endorsed the initiative through the Pontifical Academy for Life. The Pope was not a signatory in the technical sense; the document is a multi-stakeholder agreement that the Pontifical Academy signed on the Vatican's behalf.

Primary source link: romecall.org (the initiative's official website). The original 2020 text is also archived as a PDF on vatican.va.

The six principles

The heart of the Rome Call is six principles. Each is stated in a single sentence in the original document. Below is each principle, plus the practical implications the document and its commentators have drawn from it.

1. Transparency

"AI systems must be explainable."

AI cannot operate as a black box, especially when it influences decisions about human beings. People affected by AI-driven decisions should be able to understand, at least at a meaningful level, how those decisions were reached. This connects to a broader call in the document for a "duty of explanation," echoing the European Union's General Data Protection Regulation (GDPR), which gives those subject to automated decisions a right to "meaningful information about the logic involved."

2. Inclusion

"The needs of all human beings must be taken into consideration so that everyone can benefit and all individuals can be offered the best possible conditions to express themselves and develop."

AI must not discriminate, must not exclude, and must serve the underprivileged as well as the privileged. The Rome Call is explicit that the dignity of every human being grounds this principle. Inclusion has both a technical dimension (AI training data must not exclude or marginalize groups) and a social dimension (the benefits of AI must accrue to all of humanity, not concentrate in the hands of those who control the technology).

3. Responsibility

"Those who design and deploy the use of AI must proceed with responsibility and transparency."

There must always be a human being who takes responsibility for what a machine does. The Rome Call rejects the evasion of "the algorithm did it." AI systems are not moral agents; the responsibility for their actions belongs to designers, deployers, and users. This principle is foundational because it preserves human moral agency in an age when it might otherwise be diffused into systems and processes.

4. Impartiality

"Do not create or act according to bias, thus safeguarding fairness and human dignity."

AI systems must not follow biases or create them. The Rome Call recognizes that bias in AI systems often reflects bias in the data they are trained on, but argues this does not excuse it. Designers have a positive responsibility to audit AI for bias and to correct it. This principle is particularly urgent in domains where AI affects opportunity, including hiring, lending, criminal justice, and healthcare.

5. Reliability

"AI systems must be able to work reliably."

AI must do what it claims to do, and it must do so consistently. Systems that fail in unpredictable ways, that hallucinate, or that produce different outputs under indistinguishable inputs cannot be trusted with consequential decisions. The Rome Call's principle of reliability connects to a broader argument that AI must earn the trust placed in it through demonstrated performance.

6. Security and privacy

"AI systems must work securely and respect the privacy of users."

AI systems handle sensitive information and operate in environments where security failures have real consequences. They must protect the data they process from unauthorized access, and they must respect the privacy of the individuals whose data they use. The Rome Call argues that privacy is not a side concern but a constitutive element of human dignity in the digital age.

What "algorethics" means

The Rome Call introduces a term that has become widely used in Catholic and broader discussions of AI ethics: algorethics. The word is a coinage by Father Paolo Benanti, an Italian Franciscan theologian who became scientific director of the RenAIssance Foundation and a key Vatican voice on AI policy.

Algorethics means building ethics into algorithms from the design stage. The contrast is with two more common approaches. One is post-hoc ethical review, where ethics is a compliance layer applied after AI systems are built. The other is external regulation, where ethics is enforced from outside the development process by legal authority.

Neither approach is sufficient, the Rome Call argues. Post-hoc review cannot meaningfully constrain systems already deployed at scale. External regulation, however essential, always lags the technology and depends on the willingness of regulators to act. Algorethics insists that ethical considerations must be embedded in the technical practice of AI development itself, in the choices made by engineers, designers, and product managers as they build.

This is not a vague aspiration. It implies specific practices: ethics review at design stage, diverse perspectives on development teams, ongoing testing for bias and harm, transparent documentation of system behavior, and a culture in which engineers feel empowered to raise ethical concerns about systems they are building. Several signatories, including IBM and Microsoft, have published internal AI ethics frameworks that operationalize the concept.

The signatory coalition

The Rome Call is unusual among AI ethics documents in the breadth of its coalition. Most ethics statements come from a single sector: technology companies, governments, academic bodies, or religious institutions. The Rome Call deliberately spans all four.

The 2020 inaugural signatories represented the four sectors in microcosm: the Pontifical Academy for Life (religion), Microsoft and IBM (technology), the FAO (intergovernmental), and the Italian Ministry of Innovation (national government).

The 2023 Abrahamic Rome Call. On January 10, 2023, leaders from the three Abrahamic faiths gathered at the Vatican to expand the initiative. Rabbi Eliezer Simha Weisz of the Chief Rabbinate of Israel and Sheikh Abdallah bin Bayyah, president of the Forum for Promoting Peace in Muslim Societies, signed alongside Catholic representatives. The expansion signaled that the Rome Call's principles are not specifically Catholic; they are framed in language that can be shared across religious traditions, and increasingly across non-religious traditions as well.

2024: Cisco and continued expansion. In April 2024, Cisco became one of the major technology companies to add its signature. The signatory list has continued to grow through technology companies, universities, and faith communities, coordinated by the RenAIssance Foundation, which serves as the institutional vehicle for the Rome Call's ongoing work.

This breadth is a feature, not an accident. The Rome Call's signers argue that AI ethics cannot be the province of a single sector. Technology companies set technical standards; governments set legal rules; faith and academic communities articulate moral frameworks. The Rome Call is an attempt to align these four communities around a shared minimum, with the expectation that each sector will go further in its own way.

The RenAIssance Foundation

In April 2021, Pope Francis established the RenAIssance Foundation at the Pontifical Academy for Life as the institutional home of the Rome Call. The Foundation has three roles.

First, it coordinates ongoing signatory work. Once an organization signs the Rome Call, the Foundation maintains the relationship, supports implementation, and provides a venue for shared learning.

Second, it conducts research and education. The Foundation funds work on AI ethics in academic and policy settings, including a network of universities incorporating Rome Call principles into their curricula. The aim is to produce a new generation of technologists trained to think about ethical questions as integral to their technical work.

Third, it serves as a Vatican voice on AI policy. Father Paolo Benanti, the Foundation's scientific director, has become one of the most cited Catholic experts on AI in secular policy settings, advising governments, international bodies, and technology companies. He served as an advisor to the Italian government on the G7 declaration on AI in 2024 and is a regular contributor to public debates on AI regulation.

What the Rome Call has accomplished

Five years after its release, the Rome Call's record is mixed in the way that the records of voluntary multi-stakeholder commitments usually are.

What it has done well: it has built a coalition that did not previously exist. The conversation among Microsoft, IBM, FAO, and the Vatican about AI ethics is now an ongoing one, with structure, momentum, and personalities. The Rome Call gave that conversation a starting frame and a continuing institution. Other major AI ethics frameworks (the OECD Principles, the EU AI Act's foundational principles, UNESCO's Recommendation on the Ethics of AI) reflect similar substance, partly because the Rome Call's signatories influenced them.

It has also legitimized faith voices in AI policy debate. Before 2020, religious participation in AI ethics was often dismissed as either irrelevant or backward-looking. The Rome Call's coalition demonstrated that religious institutions could engage AI as a technical and ethical question at a sophisticated level, and could be useful interlocutors for the technologists actually building these systems.

What it has not done: it has not constrained AI development in concrete, measurable ways. The signatories' AI products and practices are not demonstrably different from those of non-signatories. Critics like Thomas Macaulay have argued that voluntary commitments without enforcement allow companies to "pretend to be interested in following ethical guidelines and use their support of them to deflect criticism and ward off government regulation." This criticism has merit. The Rome Call is symbolic, and symbols matter, but symbols are not regulations.

The Rome Call's own architects know this. Father Benanti has said that "the best outcome will be when the Call is not needed anymore, when it can disappear and no-one notices it. That will mean it has worked." The Call's purpose is to seed a culture, not to substitute for the legal and technical work that culture will eventually demand.

How the Rome Call relates to Antiqua et Nova

The Vatican now has two foundational documents on AI: the Rome Call (2020) and Antiqua et Nova (2025). They do different things, and a reader new to Catholic AI ethics should understand the division of labor.

The Rome Call is the practical, multi-stakeholder layer. It speaks the language of policy and engineering. Its principles can be operationalized by an engineering team or written into a procurement standard. Its signatory coalition includes companies and governments, not only religious bodies. It is designed to influence how AI is actually built and deployed.

Antiqua et Nova is the doctrinal, philosophical layer. It speaks the language of theology and metaphysics. It addresses questions like what intelligence is, what the soul is, and whether AI can stand in relationship to humans. Its audience is the Catholic faithful, theologians, and anyone interested in the philosophical foundations of Catholic AI ethics. It does not give engineering guidance; it gives the conceptual framework on which engineering guidance can stand.

The two documents are complementary. The Rome Call's six principles can be derived from the deeper philosophical commitments expressed in Antiqua et Nova. The 2025 document, for its part, explicitly engages the broader AI ethics conversation that the Rome Call helped create. Pope Leo XIV, in turn, draws on both. See the compendium of his statements.

What to read first

If you want to engage the Rome Call seriously, three reading recommendations:

Read the original document. It is short, roughly 1,000 words, and available in full at romecall.org. The whole point is that the document was designed to be read by busy people, including technology executives.

Read Paolo Benanti. His writing on algorethics is the most accessible deep treatment of what the Rome Call's principles actually require in practice. His commentary explains why the document is short (so it can be read), why the principles are framed positively rather than as prohibitions (to invite engagement rather than defense), and what algorethics means as a working method.

Read Antiqua et Nova alongside it. The doctrinal note explains the philosophical commitments that make the Rome Call's principles intelligible from the Catholic perspective. Together they form a complete picture of Catholic AI ethics circa 2025.