Use of Artificial Intelligence (AI) at Civio

The present code of conduct will establish the general guidelines governing the use of Artificial Intelligence in Civio’s work. This document is the result of lengthy reflections and discussions among the organisation’s staff, board members and trustees.

We are aware of the benefits of this technology, as well as the biases and other issues it can entail. For this reason, we wish to proceed with caution. This document is a first version; we do not know how AI will develop in the future, nor all of the consequences it will have, so it is highly likely that this code of conduct will be updated in the future. We are studying and experimenting, and if a new use of AI arises that is not covered here – or if the wider context changes – there will be new discussions, and any decision will be published with complete transparency.

In general terms, AI use at Civio is rooted in three principles, which are applicable to all areas of work:

AI does nothing on its own. When it is used, it must always undergo exhaustive human review, as set out in this document. Team members always have the final say. Our successes and failures are our own. All AI use must have a clearly identified point person.
Our AI use is transparent. If AI is used in a study (in any of the hypothetical cases considered in this document) it will be declared as part of the methodology. If AI has played a significant role in any other type of document we publish, or its involvement in any other process requires explanation, this will be explained as clearly and as accessibly as possible.
Not all AI is the same. Civio only permits the use of AI systems approved by the organisation. This approval is based on ethics-based criteria, and absolute respect for personal data protection and European AI regulations.

Civio places firm limits on situations where editorial independence, individual rights, confidentiality and public trust are at stake. We adopt a flexible, reviewable and risk-based governance approach for practical or emerging uses of AI.

Use and limitations of AI, by area of work:

Creating journalistic content

All of our articles, from start to finish, are written by a human. If we have doubts about how to use a particular word or clearly express a certain idea, we ask the rest of the team. We rely on collective, not artificial intelligence. Civio’s journalistic work is carefully crafted and thoroughly researched, and we want it to stay that way. Additionally:

AI is not used, even in a consulting capacity, to make editorial decisions or define editorial priorities. What we research and how we focus our research is for the editorial team, and the editorial team alone, to decide.
AI-generated summaries will never be used as a source of information. We use direct, rigorous sources, and only allow the use of specialised AI tools that provide the original source so it can be verified. Moreover, these will only be used to search for something specific (for example, to search for the name of a particular law that regulates a particular area, but not for more general information on jurisdiction).
We never use AI in verification and fact-checking processes. In addition to the author’s own work, our team edits, reviews and checks the facts and figures included in any article before it is published.
We do not use AI to create or edit photos or videos that appear in Civio’s articles or on our social media channels. We don’t falsify reality.
Likewise, we do not, under any circumstances, use generative AI to produce images, illustrations, collages or infographics. All creative and visual work is our own, and produced in-house. There are two reasons for this decision. First, we have very high standards when it comes to the quality of this kind of content. Second, according to our position on copyright and royalties, we believe it is unethical to benefit from tools that are trained on work done by third parties without their consent or compensation.
We also do not use AI in the translation of our articles to be published on our webpage. Indeed, many of the English versions of our articles are subject to language-specific editing after being professionally translated.
Alternative text descriptions for images and graphics, which improve the accessibility of the information we share, are treated as editorial content. For this reason, they are written by a member of our team, without AI.

Internal working processes

When analysing a database, we may use AI as a tool to convert formats, or to clean or extract data. We may, for instance, convert a PDF into CSV in order to work with it more easily, but always with human supervision and review.
We do not use AI for data analysis, i.e. to ask what elements of a dataset are significant or newsworthy. As a rule, this is something our team carries out manually. Nevertheless, if we are conducting an initial exploration of data, we may consult an AI platform with prompts such as: “Write a function in R using tidyverse to calculate annual trends in the number of ambassadors by country and gender. Include counts, totals and percentages, pad with zeros and clean unknown values.”
We may also use AI to transcribe interviews, for internal use and always carefully reviewed. Special attention is paid to parts that will be used in text – these are listened to by a human. Transcription is done with a localised AI program, both to protect the interviewee’s privacy and to ensure the necessary confidentiality of our journalistic work.
We do not use AI for any phase of the recruitment and hiring process. If a person has spent time preparing their application for a position, it deserves the attention and response of another human being.

Programming

This is the area where AI makes the biggest impact. Even so, Civio has a clear boundary on using AI in its programming: AI is not used to create our applications or graphics. We use it as a programming assistant to speed up processes, resolve doubts and review code – just as we would previously have done on programming forums – but the final product is the responsibility of team members.

Communication

Everything that reaches your inbox from Civio, or that is published on our research channels, is written and managed by members of staff. This includes newsletters, press releases and social media posts.
Enquiries and questions from readers or members will never be handled by AI. It will always be a member of our team who responds to them.
We never use AI to communicate directly with readers or members. We do use automated confirmation messages in particular processes – for example, when a membership fee payment has failed, or to send donation certificates – but these are not AI-generated. If anyone writes to us with queries, complaints, proposals, to cancel a membership, or for any other reason, the response will always come from a person writing on the organisation’s behalf, not AI.
On social media, we may use AI assistants, but only for the sole purpose of adapting marketing content or other text to different platforms in order to reach the largest number of people possible. The original content is, without exception, created by a person, and the final result is and will always be reviewed by a member of our team.

Member management

We may use AI to automate certain internal tasks, to assist in managing memberships, to cross-reference databases, or for general questions on specific groups of people. We may, for instance, ask AI for a list of our members in Cádiz and the surrounding area if we were planning an event there. But we do have clear limits:

The protection of our members’ data is one of our highest priorities. For this reason, members’ personal data will only be handled locally; it is not published or shared with any external platform. Nobody outside the organisation has access to it, only Civio staff members who are designated to carry out certain tasks, such as collecting membership fees and sending newsletters. Our administrators and auditors can only access this information for the sole – and strictly controlled – purposes of financial management and fulfilling our obligations with the Spanish Tax Agency. Every entry or query to our member database is tracked, meaning access is tightly monitored.
If we use AI to ask about certain details that members and subscribers share with us, it is always in bulk. This could be, for instance, to find out how many people became members after a particular publicity campaign. But we will never create personalised profiles under any circumstances, including for marketing purposes (see points below).

Marketing

We will never use AI to profile individuals for marketing purposes. We segment data ethically and reasonably, according to different groups (subscribers, members, ex-members, occasional donors, and so on), without the use of AI assistants. We also do not use it to build individual profiles, to send hyper-personalised communications, or to infer personal motivations, personality traits or sensitive attributes.
AI is used, if at all, as an assistant, stylistic auditor or support for creating marketing content. We may use it to improve a text’s clarity and structure, but the original and final texts must both be written by a human. Civio products in which AI can be used, solely to assist in human-led review, are: promotional materials, forms, advertisements, product messages, and welcome and support materials for campaigns. It may also be used to create variations in text for customer acquisition and retention, and for claims, calls to action, Google Ads campaigns and message editing. Additionally, we may use it to improve our understanding of aggregate results from web traffic data and campaigns (visits, opens, clicks, conversions). Nevertheless, a human will always draw interpretations and make the final decisions.

First version of this document: March 2026

Should you have any doubts or queries regarding this code of conduct, don’t hesitate to write to us. We welcome feedback, and you will be helping us to improve.