在Bloomberg报导中，Lumen Technologies Inc. 执行长 Kate Johnson 表示，AI bots 现在已占全球网际网路流量超过50%（大于五成），她认为这一比例「Remarkable」且具代表性，因为多数企业的 AI 布局仍在起步阶段；她警告各产业高阶主管必须为 AI 驱动流量在资料强度、传输速度与流量规模上快速、难预测的变化作准备，并指出若目前已超过半数，未来一、三、五年可能更高。

Johnson主张，依赖运算与资料处理的企业不能再以固定流量与路径进行传统网路容量采购，而应转向以消耗量为核心的模式。她补充，自2022年11月接任Lumen（同月OpenAI发布ChatGPT）以来，该公司目前约掌握全球网际网路流量的65%。在其领导下，Lumen以5.75亿美元（$5.75 billion，约5.75×10^9美元）出售消费者FTTH业务给AT&T Inc.，并用所得清偿4.8亿美元（$4.8 billion，约4.8×10^9美元）债务；她目标在2027年底前节省10亿美元（$1 billion，约1×10^9美元）。

Lumen与 Palantir Technologies Inc. 的2亿美元（$200 million，约2×10^8美元）软体合作，据说可加强对客户规模与需求的可见度，并指引下一个合理商机。该公司已在营运中导入多个大型语言模型用于威胁识别，于2024年底侦测到与中国骇客团体 Salt Typhoon 相关的活动并驱逐疑似攻击者。Johnson认为，像 Anthropic PBC 的 Mythos 这类AI能力将进一步提升反制效率，因为威胁行为者可在纪录短时间内搜集漏洞情报，企业必须建立同等或更优的AI侦测与封堵能力。

According to a Bloomberg report, Kate Johnson, chief executive officer of Lumen Technologies Inc., said that AI bots now account for over 50% of global internet traffic, calling it remarkable because most firms are still early in their AI journey. She warned executives across sectors to prepare for AI-driven traffic patterns that are changing rapidly in intensity, speed, and volume, and noted that if AI traffic is already above half today, much higher shares are plausible in one, three, and five years.

Johnson said companies in compute- and data-intensive sectors cannot keep relying on static network-capacity planning based on fixed volumes and routes, and should shift toward consumption-based models. She added that she became Lumen’s CEO in November 2022, the same month OpenAI launched ChatGPT, while Lumen is estimated to carry about 65% of global internet traffic. Since then, Lumen sold its consumer fiber-to-the-home business to AT&T Inc. for $5.75 billion (5.75×10^9 USD) and used the proceeds to repay $4.8 billion (4.8×10^9 USD) in debt; she is targeting $1 billion (1×10^9 USD) in savings by end-2027.

Lumen also announced a $200 million (2×10^8 USD) software partnership with Palantir Technologies Inc. to improve visibility into customer scale and needs and to identify its next logical opportunity. The company has incorporated multiple large language models into network operations to detect threats, and in late 2024 it identified activity tied to the Chinese group Salt Typhoon in U.S. telecommunications networks, then ejected suspected perpetrators. Johnson argued that models like Anthropic PBC’s Mythos will further improve response, because threat actors can gather vulnerability intelligence in record time, so defenders need equal or better AI capabilities to observe, block, and counter those threats.