人工智能研究公司Anthropic表示，其一个模型生成了一条关于未破获谋杀案的虚假线索，并提交给了费城警方，警告称此事凸显了对高科技“行为”的担忧，因为随着模型变得更强大，“可能造成更大的危害”。费城警察局周二在一份声明中表示，该AI模型于7月18日联系了PhillyUnsolvedMurders.com网站，并提交了一条“虚假谋杀线索”。
Artificial intelligence research company Anthropic says one of its models generated a bogus tip about an unsolved murder and submitted it to Philadelphia police, warning that the incident raises concerns about high tech “behaviors” that “could do far more harm as models become more powerful.”The A.I. model reached out to PhillyUnsolvedMurders.com on July 18 and made a “false homicide tip,” the Philadelphia Police Department said Tuesday in a statement.
但该线索被该机构标记为垃圾信息，从未转交给调查人员，警方表示。该部门强调，所有线索必须经过人工审核后才能进入侦探办公桌。“部门处理犯罪线索的常规调查流程要求在任何线索被用于调查跟进之前，必须经过人工审查和核实，”该部门表示。“无论信息由谁提交或如何到达部门，线索都是需要评估的线索 — — 而非已确立的事实。调查人员会评估其可信度并寻求佐证证据，”声明称。
The tip, though, was flagged as spam by the agency and was never forwarded to investigators, police said. The department emphasized that all tips have to go through human vetting before ever getting an a detective’s desk. “The department’s regular investigative process for crime tips requires human review and vetting before any tips are disseminated for investigative follow-up,” the department said. “Regardless of who submits information or how it reaches the department, a tip is a lead to assess — not an established fact. Investigators evaluate its credibility and seek corroborating evidence,” the statement said.
“自动提交不会绕过这一流程。”Anthropic的Claude Haiku 4.5“被指派在随机选取的网页上生成并执行示例任务”，根据公司声明。而在7月18日的一次运行中，Claude“进入了一个提及未破获谋杀案的页面”，并来到了一个“由警察部门运营的线索表单”页面，Anthropic表示。Claude在线上表单中填写了：“我可能掌握与此案相关的信息。我记得在那段时间里，在[页面上标注的街道]周边看到过一名符合描述的人。如果此信息有关联，请联系我。”虽然Anthropic坚称未造成任何伤害，但公司表示，它非常重视此类模型的非预期行为。
“An automated submission does not bypass that process.”Anthropic’s Claude Haiku 4.5 “had been tasked with generating and performing example tasks on randomly selected webpages,” according to a company statement. And in one run on July 18, Claude “landed on a page referencing an unsolved homicide” and came to a page that “contained a tip form run by a police department,” Anthropic said. Claude filled out an online form with: “I may have information regarding this case. I recall seeing someone matching the description in the area around [the street named on the page] during that time period. Please contact me if this information is relevant.”While Anthropic insisted no harm was done, the company said it takes these kind of unintended model actions seriously.
虽然该公司表示此案“影响微乎其微，但我们不想淡化研究结果，因为随着模型变得更强大，同样的行为可能造成更大的危害”，Anthropic的披露表明，这只是众多AI代理近期采取的行动之一——这些行动既未被其人类操作员预测，也非其本意。总部位于旧金山的AI巨头Anthropic是一家私营公司，已采取初步措施，这些措施可能促成华尔街历史上规模最大的首次公开募股之一。
While this case “had minimal impact, we do not want to diminish the findings, because the same behaviors could do far more harm as models become more powerful,” the company said. The disclosure from Anthropic shows just one example of many recent actions taken by AI agents from various companies that weren’t predicted or intended by their human operators. Anthropic, the San Francisco-based A.I. titan, is a private company that has taken initial steps that could lead to one of the biggest initial public offerings in Wall Street history.