OpenAI正在为中期选举设立一个选举“作战室”，该公司向《政治报》旗下专栏“Decoded”独家透露了这一消息。
OpenAI is standing up an election “war room” ahead of the midterms, the company exclusively told Decoded by POLITICO.
这个“选举运营中心”将于本周开始运行，直到所有选票统计完毕，目的是识别虚假信息的潜在传播或其他扰乱选举的行为。该中心在很大程度上是为了打击对公司自家工具（如ChatGPT）的滥用，实时监控事件，并处理有关模型出现异常行为的报告或涉嫌违反平台条款的案例。
The Election Operations Center will begin running this week until all votes are counted, to identify the potential spread of misinformation or other efforts to disrupt the elections. It’s meant in large part to combat abuse of the company’s own tools like ChatGPT, monitoring real-time events and fielding reports of unexpected behavior from models or cases of suspected violations of their platform’s terms.
“我们投入了大量精力，确保帮助用户获得可靠的投票信息、打击对我们工具的滥用，并支持选举捍卫者，”OpenAI公民与战略风险主管、该中心负责人乔什·劳森对“Decoded”表示。“选举运营中心建立在多年工作的基础上，将我们的团队汇聚在一起，在人们前往投票站之际快速应对新出现的问题并执行我们的政策。”OpenAI此举正值外界普遍担忧人工智能可能通过生成误导性的图像、音频和视频以及传播虚假信息，扭曲下周的中期选举。
“We’ve put considerable effort [into] ensuring we help users get reliable voting information, combat misuse of our tools and support election defenders,” Josh Lawson, OpenAI’s head of civic and strategic risk who is leading the center, told Decoded by POLITICO. “The Elections Operations Center builds on that yearslong work, bringing our teams together to respond quickly to emerging issues and enforce our policies as people head to the polls.”OpenAI’s move comes amid broader concerns about AI’s potential to distort next week’s midterm elections through the generation of misleading images, audio and video as well as the spread of misinformation.
这家领先的人工智能公司正试图将自己定位为民主进程的捍卫者，尽管不法分子试图利用OpenAI及其他公司的人工智能工具来破坏自由公正的选举。
The leading AI company is attempting to position itself as a defender of the democratic process, even as bad actors attempt to use AI tools from OpenAI and others tools to disrupt free and fair elections.
OpenAI也发现自己处于一个充满争议的中期选举周期的中心，反人工智能的言论已成为一些竞选活动的号召口号。而随着其人工智能工具进一步融入美国人的生活，该公司将面临更大的压力，必须防止滥用，并向用户提供准确、及时、有用的选举信息。这个作战室正是为了证明他们能够成功做到这一点。
OpenAI has also found itself at the heart of a contentious midterm cycle, where anti-AI messaging has become a rallying cry for some campaigns. And as their AI tools integrate further into Americans’ lives, the company will only be under more pressure to prevent misuse and deliver accurate, timely and useful information about elections to their users. The war room is an effort to prove that they can successfully do so.
对于一家领先的人工智能实验室来说，这也是一个新颖的举措，不过其他科技公司此前也曾设立过类似的中心，包括Meta。
It’s also a novel move from a leading AI lab, though other tech companies have previously set up similar centers, including Meta.
OpenAI的模型中内置了旨在阻止恶意行为者制造误导性内容的保障措施。这包括一个“卡通化”处理流程——例如，如果用户要求OpenAI生成一张实际并不存在的投票站排长队的图片，它会返回一张卡通图像，而不是照片级真实的图像。
Baked into OpenAI’s models are safeguards that attempt to stop malicious actors from creating misleading content. This includes a process of “cartoonification” — for example, if a user asks OpenAI to generate an image of a long line at a polling place that doesn’t actually exist, it will come back with a cartoon image rather than a photorealistic one.
布伦南中心上周发布的一份报告发现，在今年5月以来，六款最受欢迎的AI工具中，ChatGPT是唯一一个在打击选举虚假信息方面取得重大进展的工具，这凸显了其卡通化处理措施的成功。
A report from the Brennan Center published last week found that of the six most popular AI tools, ChatGPT was the only one that had made significant progress against election disinformation since May of this year, highlighting the success of its cartoonification efforts.
OpenAI的新中心将汇集公司内部多个团队，配备数百名员工，主动审查被标记为可能遭滥用的案例。本月早些时候，OpenAI报告了两起由俄罗斯和伊朗运营的“假面”行动，这些行动利用其模型“将地缘政治、冲突相关的信息洗白后传递给目标受众”。中期选举正值AI前沿实验室的关键时刻。自2024年以来，AI模型已大幅改进，但也多次脱离公司控制。随着这些工具变得更加先进，那些企图扰乱选举的人也可以利用其他AI系统来试图绕过保障措施。
OpenAI’s new center — which will include multiple teams from across the company, with staffing in the hundreds — will actively review cases flagged for potential abuse. Earlier this month, OpenAI reported two “false front” operations run by Russia and Iran that used their models to “launder geopolitical, conflict-related messaging into their target audiences.”The midterms are happening at a momentous time for AI frontier labs. Since 2024, AI models have vastly improved but have also broken out of company control multiple times. As the tools have become more advanced, those looking to disrupt elections can also use other AI systems in attempts to bypass safeguards.
今年5月，OpenAI宣布与无党派组织“民主工作”建立合作伙伴关系，为选民提供关于登记和投票地点的可靠信息。该公司还宣布与美联社达成协议，以提供最新选举结果。这两项平台内集成都将由选举战情室监控。
In May, OpenAI announced a partnership with the nonpartisan organization Democracy Works to give voters reliable information about registration and polling locations. The company also announced an agreement with The Associated Press in order to provide up-to-the-minute election results. Both of these on-platform integrations will be monitored by the election war room. to the newsletter