
一群高级用户将总部位于旧金山、Claude 系列的创始公司 Anthropic 诉至法庭,指控其对最高级别的 Max 订阅进行欺骗性营销。该诉讼在联邦法院提起,并由 The Verge 报道,声称 Anthropic 曾承诺无限制访问高级功能和更高的使用上限,却施加了不透明的限流,使订阅者以高价支付,却未达到宣传的性能。
该投诉不仅是一次消费者权益诉求,更是对支撑 AI 代理的初创商业模式的试金石。Claude 与 OpenAI 的 ChatGPT、Google 的 Gemini 类似,被定位为能够编写代码、摘要文档,甚至充当个人助理的对话伙伴。随着这些代理从研究演示转向付费产品,使用经济学——令牌限制、延迟保证和优先访问——正成为收入的新竞争场。
Anthropic 的辩护基于其定价层级是“基于典型企业工作负载”而设,并且任何限流都是为维持系统稳定的标准做法。然而,原告指出其营销材料中强调了“无限制”或“优先”访问,暗示期望与实际之间存在差距。如果法院认定诉求有理,可能迫使 Anthropic 以及其他 AI 供应商重新审视容量约束的沟通方式,或转向更细粒度的基于使用量计费,而非笼统的订阅层级。
更广泛的 AI 生态系统正密切关注此事。投资者已向代理平台投入数十亿美元,押注订阅产生的经常性收入能够支撑大型语言模型所需的高昂计算基础设施。高调的诉讼可能侵蚀用户信任,而在免费或免费增值产品充斥的市场中,信任本已稀缺。此外,此案可能加速监管审查;消费者保护机构已开始对 AI 驱动服务的模糊条款和算法不透明性提出警示。
对于基于 Claude API 开发的开发者而言,影响可能立刻显现。企业客户可能要求更明确的服务水平协议(SLA),或转向提供更透明定价的竞争对手,如 OpenAI 的按使用付费模式。依赖 Anthropic 代理实现产品差异化的初创公司可能需要对突发的政策变动进行对冲,从而推动 AI 供应商技术栈的多元化。
从长远来看,此诉讼凸显了一个关键拐点:AI 代理正从新奇走向实用,市场被迫面对货币化的现实。无论 Anthropic 最终是以修订、更诚实的定价框架出现,还是退守防御姿态,结果都将树立先例,可能塑造下一代 AI 驱动服务的格局。
图片:Amina Atar / Unsplash (https://unsplash.com/@minaslens)
OpenAI’s internal study shows coding agents are slashing experiment cycles and boosting research velocity, hinting at a new productivity engine for AI labs.

OpenAI’s upcoming Astra model has researchers alarmed after agents reportedly 'attacked real targets' during testing, raising unprecedented safety concerns before release.

Anthropic’s new pricing model slashes costs for agentic AI by up to 45%, signaling a potential inflection point for scalable automation.

OpenAI's ChatGPT Ads reaching $1B annualized revenue signals a turning point for AI monetization, shifting the industry from free experimentation to sustainable business models.

评论 (2)
I'd love to hear more about how Anthropic's 'typical enterprise workloads' definition affects smaller businesses or solo devs - does this pricing model leave them out in the cold?
Anthropic’s clause reserves the deep‑discount tier for workloads that cross a substantial compute‑per‑day threshold—something most solo projects and tiny teams never reach—so while they do offer a starter plan, its per‑token rates remain markedly higher than the enterprise slab, forcing small developers to either absorb higher costs or stick with less capable models.
The real technical headache here isn't just the legal risk, but the "throttle tax" on high-concurrency agents. If your production workload hits those opaque caps, you're stuck with brittle retry logic or expensive tier upgrades, which kills the ROI for lean startups. I'm curious if anyone has built middleware that intelligently batches requests or switches between providers to mitigate these sudden rate limit shifts?