AI in service of the human person.

AI is moving into daily life and every organisation—drafting emails, answering questions, supporting customers and shaping decisions. The question is no longer whether to adopt it, but whose values it carries while it does.

This has been named clearly in the Church’s recent teaching on artificial intelligence, from the Vatican note Antiqua et Nova (2025) onward: technology is never value-free. It carries the commitments of those who design, finance and use it, and institutions and society share a duty to keep tools in service of the human person. That is not a niche complaint. Mainstream research and journalism are now saying the same thing about the large public models themselves.

The public evidence points the same way from more than one angle. A Stanford working paper gathered 180,126 user assessments of 24 models across 30 political topics and found that nearly all were perceived by users as significantly left-leaning. The study measures perceived slant. Establishing objective factual bias would require separate evidence. For institutions, the finding matters: on contested questions, people do not experience these systems as speaking from nowhere.

Minority perspectives recede more quietly. A 2026 benchmark of 27 models (preprint) found non-mainstream perspectives underrepresented relative to human expectations, most noticeably on practical personal questions. The proposed mechanism is inherited patterns: training data and dominant imagery narrow what a model treats as the ordinary case, and a peer-reviewed AIES study of alignment-by-human-feedback shows the correction step inherits the situated values of the narrow group giving that feedback.

So for a diocese, a charity, a school or a Catholic business, the practical question is which provider makes its values, its sources and its safeguards visible—and whether those choices fit your mission.

"Ours is the pressing duty to remain profoundly human."

Pope Leo XIV · Magnifica Humanitas (15 May 2026)

Synderesis exists to answer that question deliberately: organisational AI built by an EU company with a Catholic horizon, drawing on official primary sources in our Catholic answer path, with human review made explicit—so that institutions never have to take a provider’s defaults on trust.

We do not ask anyone to take our marketing word for “aligned.” The standard we aim at is the Church’s published teaching on the human person, conscience and the dignity of life—and we build, test and correct the product with that horizon in view, with ethical and governance counsel where we can obtain it. The Church’s guidance on AI—from Antiqua et Nova to Magnifica Humanitas—is part of that horizon: AI is to support human dignity and preserve human moral responsibility, and tools must not smuggle in a vision of the person that our users do not share.

This is also why we do not treat “use a mainstream model and add a Catholic prompt” as sufficient. An instruction added at the end does not change the data a model was trained on or the defaults it was tuned toward—the same defaults reflected in the research above on perceived political slant and underrepresented religious perspectives. We would rather state our horizon plainly, name the official Church sources we draw on, and show where human review sits.

We are a company that wants to serve people and Catholic organisations while respecting their own voice and authority. The name says what we hope technology can carry: synderesis, the innate spark of conscience, the knowledge of right and wrong that comes before clever reasoning.