Helping to safeguard the integrity of elections and more: overview of recent vera.ai activities
Over the past months the vera.ai project participated in a number of significant webinars and events aimed at training fact-checkers and raising awareness of civil society on deepfakes and AI manipulations. This included disseminating our latest tools, sharing best practices, and raising awareness about potential threats. In addition, the vera.ai project coordinator Symeon Papadopoulos and Nikos Sarris (both from CERTH) and Denis Teyssou participated in the deepfake task force set up prior to the EU election by EDMO and the EFCSN. This contribution provides an overview of some of these activities and events, written by one of its protagonists, Denis Teyssou.
Webinar 1: Helping the Shakti project to support Indian fact-checkers
For the first online webinar, we were invited by the Indian company DataLeads and the Shakti project on election integrity. On 28 March 2024, we shared best practices and our available tools to combat deepfakes in video, audio and AI-generated images with about 100 Indian fact-checkers.
Vera.ai innovation manager Denis Teyssou (the author of this contribution) presented the results of the tools integrated into the verification plug-in on disinformation cases reported in recent Indian fact-checks.
This included deepfake videos of politicians, business leaders or celebrities as well as generated images. Participants learned about various available verification techniques and how to effectively debunk the manipulated content. The session empowered the Indian fact-checking community by equipping them with the knowledge and tools necessary to actively contribute to the protection of electoral processes.
Following this webinar, we built a specific version of the plug-in for the EU Parliament with their logo, providing access to our latest features that are still under active development.
Webinar 2: Training the EU Parliament on deepfake detection techniques
The second webinar, at the request of the European Parliament, delved into the use of technology in protecting elections against AI manipulations at the request of the European Parliament. The online event, held on 12 April 2024, attracted around 500 participants from the institution, demonstrating a significant level of concern regarding the threat of deepfakes.
Denis Teyssou showcased the latest innovative digital tools as well as some other detectors designed to analyse AI-manipulated content. We highlighted recent case studies where technology had successfully detected and mitigated disinformation targeting politicians and public figures. This webinar underscored the importance of technological awareness to anticipate any possible disinformation campaign using AI manipulation techniques.
Webinar 3: Training European fact-checkers with EDMO
The third and final webinar of this series, held on 23 May 2024, provided an advanced EDMO training on AI manipulation detection tools integrated into the InVID-WeVerify plug-in, now supported and enhanced further by vera.ai.
Denis Teyssou provided a 90 minute training on the latest features integrated into the plug-in as beta tools reserved for fact-checkers, journalists and researchers tackling disinformation.
It included voice cloning detection, synthetic image detection and deepfake (mainly face swapping) detection. Some 50 European fact-checkers attended the workshop and were all granted access to the new beta tools.
Summing up
In conclusion, participation in these webinars was a testament to the vera.ai team's commitment to empowering fact-checkers and civil society to protect democratic processes. Through these educational and collaborative efforts, we contributed to raise awareness about AI manipulations and the tools available to counter them. The knowledge and skills imparted during the sessions gathered very positive feedback and helped us to align the features that are being developed and provided further to many different "real life" use cases.
Author: Denis Teyssou (AFP)
Editor: Jochen Spangenberg (DW)
