CNET警告:勿用AI聊天机器人决定投票
CNET报道称,在美国中期选举前,部分民众正使用ChatGPT、Claude等生成式AI聊天机器人了解候选人和提案。文章警告,聊天机器人可能给出有偏见、迎合用户、过时或虚构的信息,因此应核查引用,并通过州或地方政府官方渠道确认选举信息。CNET对四款助手的测试中,Claude明确拒绝推荐候选人,Gemini显示核验提示,而ChatGPT和Grok未作特别提示。
报道揭示了AI聊天机器人用于政治和选举研究时的实际可靠性风险,也显示不同助手在政治中立提示与来源核验方面存在差异。

There are hundreds of races happening during this year’s midterm elections, from every seat in the US House of Representatives to various state and local government roles. Many tech issues are on the ballot, including AI policies, but people are also reportedly turning to generative AI chatbots to help them prepare for Election Day.
One of the biggest uses for chatbots is searching for information, so it isn’t surprising to see people using ChatGPT, Claude or another AI tool to learn about their candidates and ballot measures. But we know that generative AI tech has some serious risks, such as bias, sycophancy, hallucinations and drawing incorrect conclusions using bad information.
But if you’re going to use chatbots for your election research, make sure you take these precautions. You can also check out CNET’s guide to spotting political scams and deepfakes.