AI companies promise to protect our elections. Will they live up to their pledges?

Observers warn that major AI firms have not translated election integrity pledges into concrete, transparent safeguards, urging safe-harbor policies for independent research, regular public reporting, multilingual capacity, and stronger provenance tools. Companies cite steps like watermarking (Google’s SynthID), red-teaming, and partnerships with election authorities, but experts say measures are nascent and unenforceable; studies show chat models still provide inaccurate voting information, and firms are delaying broad release of powerful video generators (OpenAI’s Sora, Google’s Veo) amid misuse risks.