WITNESS Launches Its First Regional Deepfake Rapid Response Force in Brazil
The new initiative brings together Brazilian fact-checkers and forensic experts with WITNESS’ expertise to detect AI-manipulated media before it can distort the 2026 election in the country.
Read this text in Portuguese here.
To safeguard Brazil’s 2026 elections, taking place in October, WITNESS is adapting its Deepfakes Rapid Response Force (DRRF) model for the country’s fast-moving information landscape. Since 2023, WITNESS has led the first initiative of its kind globally: a Deepfakes Rapid Response Force connecting frontline journalists, human rights defenders, and fact-checkers in real time with leading AI detection and media forensics experts. The global Force has helped identify and expose synthetic media in high-stakes moments around the world, protecting the integrity of information when it matters most.
Earlier this year, Brazilian fact-checking organization Aos Fatos identified at least 50 TikTok videos simulating street interviews and testimonials about Brazil’s leading presidential candidates, Lula and Flávio Bolsonaro. None of the “voters” in them were real. The videos had accumulated millions of views, illustrating the scale and speed at which synthetic content can spread during an election.
The Brazil DRRF marks the first time WITNESS has adapted this model at a regional level. Developed in collaboration with Brazilian fact-checking organizations Aos Fatos, Agência Lupa, and Projeto Comprova, and Recod.ai, a research group at the University of Campinas (Unicamp), the Force will give investigators and reporters a secure channel to submit suspicious audio, video, or imagery for rapid analysis, combining local journalistic expertise with specialized AI detection and forensic tools.
Deepfakes and the uncertainty about what we can trust
“Deepfakes don’t just deceive people at the moment. They create lasting uncertainty about what we can trust, making it easier for fake content to be believed and real evidence to be dismissed,” says shirin anlen, AI Research and Technology Manager at WITNESS.
“The Local Deepfakes Rapid Response Force gives Brazilian journalists and fact-checkers rapid access to reliable expertise to navigate that uncertainty with confidence and credibility. And what we learn from this election will help strengthen how we respond to these threats well beyond Brazil,” anlen adds.
“As artificial intelligence algorithms continue to advance, ensuring the integrity of content circulating in the media has become an increasingly complex technical challenge — and a fundamental requirement for the integrity of electoral processes. At the same time, AI itself offers an important part of the solution, enabling the development of increasingly sophisticated methods to identify manipulated content and protect the authenticity of information,” says Michael Macedo Diniz, Research Collaborator at Recod.ai.
“For us, being part of this initiative is an opportunity to make a concrete contribution to protecting elections and democracy, while also bringing our research into contact with real-world challenges and cases. This allows us to identify emerging threats and develop solutions capable of keeping pace with the rapid evolution of these technologies,” Diniz adds.
The launch reflects WITNESS’s broader push to build regional capacity for media forensic analysis and AI detection — strengthening local ecosystems rather than centralizing expertise, and positioning trusted local institutions to respond to synthetic media threats as they emerge. WITNESS plans to draw lessons from the Brazil pilot to inform future regional deployments of the Force elsewhere.
Media Contact: Jéssica Ribeiro, Senior Communications & Engagement Coordinator, WITNESS| Jessica@witness.org|