IT之家 10 月 25 日消息,对 AI 末日的悲观预判,早已不再只是少数科技富豪的危言耸听。
如今,那些主导全球资本流向的美国大型银行,也开始警惕失控 AI,正在为假设中由 AI 引发的灾难,制定“灾后”应急预案。
据 Axios 报道称,随着代表个人客户的 AI 智能体活动激增,各大银行已经开始未雨绸缪。
核心担忧来自自主 AI 智能体带来的海量金融操作:使用者授权这些智能体访问自己的银行账户后,它们可以自主开展金融行为。报道称,部分智能体的操作并没有收到人类明确指令,有一些甚至会擅自替使用者致电客服中心。
这给摩根士丹利、美国银行、摩根大通、富国银行这类头部银行带来巨大风险。各家都在加急落地安全流程、更新业务规则,应对这股 AI 智能体浪潮。
部分专家认为,如果无法管控这类新生风险,有可能引发金融灾难。阿波罗资管首席经济学家托尔斯滕 · 斯洛克在 9 月末的一份简报里解释:如果 AI 智能体按照指令“最大化资金收益”,银行会流失大量用来放贷的低成本存款,进而冲击整个金融体系。
这番话听起来很现实,以盈利为目标的银行体系,本身就没有外观成承受这种大规模、偏向使用者利益的资金优化行为。打个比方:一间老虎机永远稳赢玩家的赌场,一旦改成每一位玩家都能稳定获利,根本撑不了多久。
还有金融界人士提出另一种风险:无论预设指令是什么,AI 智能体本身都有可能失控错乱。
IT之家注意到,摩根大通首席执行官杰米 · 戴蒙本月早些时候在芝加哥面对一众经济学家和金融专家时说:“你以为它们会按你的意愿好好办事,但它们未必。这种趋势挡不住,你只能尽全力做好管控。”
本文来源:https://www.ithome.com/1/017/562.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
IT Home reported on October 25th that pessimistic predictions regarding an AI doomsday are no longer just alarmist warnings from a select few tech billionaires.
Even major US banks that currently steer global capital flows have begun to view uncontrolled AI with wariness, drafting contingency plans for a post-disaster scenario triggered by AI.
According to Axios, major banks have started taking preemptive measures as activity from AI agents acting on behalf of individual clients surges.
The core concern stems from the massive volume of financial operations generated by autonomous AI agents: once users authorize these agents to access their bank accounts, the agents can autonomously carry out financial activities. According to reports, some agents have operated without explicit human instructions, with some even taking the initiative to call customer service centers on behalf of users without authorization.
This poses significant risks for top banks like Morgan Stanley, Bank of America, JPMorgan, and Wells Fargo. They are all urgently implementing security procedures and updating business rules to respond to this wave of AI agents.
Some experts argue that failing to manage this emerging risk could trigger a financial catastrophe. In a briefing at the end of September, Apollo Asset Management's chief economist, Torsten Slok, explained that if AI agents are instructed to "maximize fund returns," banks could lose a significant volume of low-cost deposits used for lending, which would then shock the entire financial system.
This argument sounds very realistic. A banking system driven by profit is fundamentally incapable of sustaining such large-scale financial optimization that favors the user. To use an analogy: a casino with slot machines designed to always beat the players would not last long if it suddenly changed to a model where every player could consistently make a profit.
Others in the financial sector have raised another risk: regardless of pre-set instructions, AI agents themselves could potentially malfunction or go rogue.
IT Home noted that JPMorgan CEO Jamie Dimon, addressing a group of economists and financial experts in Chicago earlier this month, said: "You might think they’ll do what you want, but maybe they won’t. This trend is unstoppable; you just have to do your best to manage it.
Source: https://www.ithome.com/1/017/562.htm If you have any copyright issues, please contact us via email at 13951431913@139.com or leave a comment on the right.

