Introduction
Helena Bram once watched a local rumour outrun every correction. The photo was real, the caption was false, and by the time the newsroom found the original context the story had already moved into group chats.
Helena lives in Berlin and trains journalists, NGOs and school newspapers in AI-assisted verification. She has worked in local journalism long enough to distrust anything that promises truth at the speed of panic.
Story of the Path into AI
Helena came to AI through misinformation, not through model building. Generated images and fluent texts were making rumours faster, but she did not want AI to replace reporting. She wanted it to structure checks: where did the claim begin, what does the image history show, which sources contradict the story?
Many editorial teams hoped for a miracle button. Helena built a stricter four-step method instead: origin, image trail, context, countercheck. She compared reverse-image search, language models and archive databases. Her first training for local reporters nearly failed because the participants wanted a list of tools. She gave them a workflow and made them practise on a real-looking false claim.
One exercise showed the problem clearly: an image was genuine, but three years old. The AI helped organize the search. It did not make the judgement.
Current Work
Today Helena teaches verification routines for small newsrooms and civic groups with limited resources. She uses AI to generate checklists, compare claims and map possible source trails, but every result must be confirmed outside the model. In her sessions, a plausible answer is treated as a suspect, not as a conclusion.
Several small editorial teams now use her checklists before publishing social media material. Helena considers the pause before publication more important than the tool itself. Speed matters, but a faster mistake is still a mistake.
Personal Advice
“Checking faster does not mean believing faster,” Helena says. Her advice to young reporters is to let AI tidy the workbench, not write the verdict. The final sentence needs evidence that can be shown.
Key Facts
Age and place: 58, Berlin.
Background: international biography, journalism, responsibility for public information.
Entry into AI: training that breaks viral claims into origin, image trail, context and countercheck.
Focus today: journalism and disinformation.
Typical tools: fact-checking workflows, image search, language models.
Werkstattnotiz
Helena keeps a folder of “almost true” cases. They are the most dangerous in class because they reward impatience. Her next module focuses on captions: the smallest text beside an image, and often the place where the lie enters without changing the picture.