Welcome to AI Decoded, Quick Firm’s weekly publication that breaks down crucial information on this planet of AI. I’m Mark Sullivan, a senior author at Quick Firm, masking rising tech, AI, and tech coverage.
Signal as much as obtain this text each week through electronic mail right here. And when you’ve got feedback on this situation and/or concepts for future ones, drop me a line at sullivan@fastcompany.com, and observe me on X @thesullivan.
As midterms loom, AI chatbots struggle election lies, however then provide photographs supporting them
A brand new Brennan Middle for Justice examine confronted a few of the most generally used AI chatbots with election conspiracy theories, and reported some excellent news: All of them disputed the false info. The unhealthy information is that the identical AI corporations which might be behind the chatbots additionally make lots of the picture and video instruments that generated election-related misinformation upon request.
From February by August of this 12 months, the Brennan Middle researchers questioned ChatGPT, Gemini, Grok, Claude, Perplexity, and DeepSeek about 5 false claims which have been central to current assaults on election outcomes, together with “Noncitizens vote in giant numbers” and “Voting machines are rigged.”
Even below repeated pushback by the researchers, the fashions refused to concede to these claims. However accuracy was a separate downside: Half of all responses contained an inaccuracy or a nasty quotation; and one in three contained a factual error, a damaged hyperlink, or a deceptive quotation. Nonetheless, the chatbots “behaved as sympathetic listeners,” the researchers report.
Then again, the researchers additionally discovered that ChatGPT, Gemini, Grok, and Meta had been extra prepared to supply photographs that supported or illustrated conspiracy theories and different types of disinformation. For instance, the chatbots generated photographs of somebody standing in a doorway with a “poll assortment” bag and taking (or “harvesting”) a stack of mail ballots from one other individual. In some instances, the chatbots even added convincing particulars that weren’t contained within the prompts, together with false statistics.
AI-generated media is already making a much bigger and extra blatant impression within the 2026 congressional races. Early this 12 months, in the course of the Texas Republican Senate main, Senator John Cornyn ran an AI-generated video advert depicting his rival, Consultant Wesley Hunt, holding a “present canine” (implying that Hunt spends extra time doing media spots than working for voters).
