OpenClaw alarms, Tencent Cloud price shock and Anthropic's lawsuit underscore AI's fraught moment

Domestic turbulence: agents set off official alarms while cloud bills jump

China's rapidly spreading AI agent ecosystem has run into regulatory headwinds. OpenClaw (龙虾), an open-agent project that has been widely adopted by firms and hobbyists, has drawn consecutive warnings from authoritative agencies — including the Ministry of Industry and Information Technology's threat-and-vulnerability sharing platform, the National Internet Emergency Response Center (CNCERT), and the National Cyber Information Center — which flagged high-security risks in default or improper deployments and catalogued hundreds of historical vulnerabilities. The agencies warned that exposed OpenClaw instances, reportedly more than 200,000 worldwide and about 23,000 within China, could become easy targets for network attacks and data leaks. Who benefits when a technology designed to automate tasks becomes a national security headache?

At the same time, Tencent Cloud (腾讯云) startled enterprise customers by announcing sweeping price changes effective March 13, 2026. Several models including GLM5, MiniMax2.5 and Kimi2.5 are exiting free public beta and moving to pay-as-you-go. Most eye-catching: Tencent's mixed-series models Tencent HY2.0Instruct and Tencent HY2.0Think saw input and output prices raised by roughly 400–460% (for example, HY2.0Instruct input from ¥0.0008 to ¥0.004505 per 1,000 tokens). The move follows a broader industry recalibration as providers seek sustainable economics for inference-heavy services; it also risks squeezing startups and developers already grappling with infrastructure and security costs.

International fallout: Anthropic fights U.S. supply-chain move

The unsettled picture is not limited to China. In the U.S., Anthropic has sued the federal government after being designated a supply-chain risk, a move that reportedly led to cancelled contracts and significant reputational and financial harm for the startup. The dispute centers on whether the company's Claude models could be repurposed for military uses; Anthropic says it will not enable fully autonomous weapons or broad surveillance, and calls the government's action unprecedented and unlawful. The U.S. Department of Defense has signaled little appetite to resume military negotiations with Anthropic, even as the company pursues legal remedies.

The three threads — insecure, widely distributed agents at home; sharply higher cloud pricing that re-prices AI usage; and cross-border legal fights over AI's military uses — point to the same tension: AI is moving from lab novelty to infrastructure and geopolitics almost overnight. Companies, regulators and governments are all recalibrating at once. Expect more friction before the margins and rules settle.