Claim: The 2026 US midterm elections will be stolen through AI deepfakes and widespread voter fraud

The claim that the 2026 US midterm elections will be stolen through AI deepfakes and widespread voter fraud is mostly false. While there are concerns about AI deepfakes potentially undermining public trust, credible sources do not support the assertion that these will lead to stolen elections. Reports from mainstream outlets highlight the risks of misinformation but do not provide evidence of actual voter fraud or election theft. Alternative sources may exaggerate these risks, suggesting a more sensational narrative without substantial proof. Overall, the evidence indicates that while AI deepfakes pose challenges, they are unlikely to result in the theft of elections. The graders agree on direction, but vary in strength. OpenAI comes in highest (20%), while Gemini is lowest (5%). Gemini expresses higher confidence than OpenAI on this claim. Opposing sources argue that AI deepfakes are becoming increasingly sophisticated and could significantly impact voter perceptions and trust in the electoral process. They highlight incidents of deepfakes in political campaigns that could mislead voters. However, these claims do not provide concrete evidence of voter fraud or election theft, which is crucial to substantiate the original claim. The lack of verified instances of stolen elections due to deepfakes suggests that while the risks are acknowledged, they do not equate to the claim of elections being stolen. Thus, the uncertainty does not alter the overall verdict of mostly false.