By Steve Shipside
“AI is already in your organization. The question here is whether you know it or you don’t.” Olalla Novoa Ojea does not mince her words when it comes to corporate control of the new generation of tools. “77% of employees copy paste internal data into AI tools that they control personally.”
Given she was speaking at the World News Media Congress session featuring lessons learned from WAN-IFRA’s Newsroom AI Catalyst initiative (which she and Prisa took part in), you might think that was good news, but she was quick to point to the problem.
“45 percent of AI responses on news contain errors. When an AI error happens in a newsroom, that’s a whole other story… credibility is the foundation of everything we build in the media.”
Scale compounds the credibility issue
Prisa operates in 12 countries with 25 brands including entities such as Los 40, Cadena SER, Cinco Dias, Caracol Radio, W Radio, and El País, so true governance becomes daunting. “You are talking about dozens of newsrooms, different countries, different legal jurisdictions.” It also meant some 30 AI projects being developed, and it was starting to look very much as if the AI footprint was fast outpacing any system designed to manage it.
Which left Olalla with a key decision: “keep going as we were, launching quick use cases, assuming the risks, or building a governance framework, stopping momentarily to make sure that we could run afterwards.”
Olalla went for the second option, prioritising structure and control over quick wins.
Why brakes don’t have to mean slowing
Braking doesn’t quite sound right in the context of an AI accelerator, but her point was that governance would mean faster real progress than careering into obstacles. “We understand governance not as a brake, but as an engine. When people know what they can do or they cannot do, then they adopt AI with more confidence.”
Use responsibly
Of course there’s a world of difference between having a policy, and making it happen. She points out that there was already a responsible AI use policy in the company, dating back to 2024 – but a management email and daily practice are not the same.
Prisa put together an AI Oversight Committee, and importantly combined not only legal and compliance experts, and the technologists, but also editorial, HR, sustainability, and leadership experts. The committee reviews every proposed AI usage, but the basis of their decision comes down to three key questions.
What, how, why?
The ‘what’ aligns perfectly with the point made by the WAN-IFRA AI accelerator stewards [link to earlier article] that however sparkly AI tools might be, they have no place in publishing unless they are solving a problem. The ‘how’ is very much about who will be responsible, and the ‘why’ takes in not only business goals, but the company’s ethical commitments, including to its own staff.
Spreading the word
It’s an industry irony that publishers aren’t always the best at communicating internally, so part of the AI governance involved creating an internal AI hub with clear explanations of policies, protocols, training, communities, and sources of information. The communities feature was particularly important, not only to give a sense of belonging at a time of change, but also to help avoid independent and unsupervised experimentation.
That’s an ongoing task, since what Olalla calls ‘shadow AI’ has not gone away and continues to be a potential risk, but structure and framework help discourage it.
Results
Over 900 people have now been trained on the responsible use of AI, and they all have that hub as a resource to go back to afterwards, plus 21 tools have been approved, and Olalla notes that “we have a catalog of any tool that is used and any project that is going on with full documentation … this was not happening before.”
Of course that’s by no means the end of the story, and Olalla is quick to point out that vibe coding is the next big challenge; “everyone can code now, but how do you integrate that within a large company in a secure way?”
Conclusion
“Trust is our most precious asset. Whenever anyone opens El País or whenever anyone tunes in to Cadena SER… they have made a choice to trust. We need to protect that. And the way to protect that in the AI world is through governance.”
Because brakes don’t have to mean slowing; they mean control, and that means prioritising progress, over raw (uncontrolled) speed.
Steve Shipside is a freelance journalist and media consultant who has covered the news industry for more than 25 years, and has contributed to WAN-IFRA’s content for several years.