August 4 was a day when AI policy direction and infrastructure investment moved at the same time. A picture emerged of major US Silicon Valley players splitting squarely down the middle over how to handle Chinese open-source AI models. Meanwhile, Anthropic signed a massive compute deal with a newly founded cloud upstart to keep up with surging demand, and CoreWeave made its first entry into the Asian market. On the security front, the reality that floods of AI-generated false reports (AI slop) are paralyzing corporate vulnerability handling was thrown into sharp relief once again.
US Silicon Valley Splits Over Chinese Open-Source AI Regulations
According to a New York Times report, the Trump administration had been discussing sanctions and cloud usage restrictions with Chinese open-source AI models in mind. Here, the industry's response is divided. OpenAI and Anthropic called for stronger regulations, while Nvidia, Google, and Meta opposed them. After pushback from Silicon Valley, the administration reportedly held off for the time being, but some decision is expected before President Xi Jinping's planned visit to the United States in September.
Regulation of open-source models intertwines two different motives: national security concerns and the competitiveness of each company's own models. The pro-regulation camp of OpenAI and Anthropic centers its business on closed models, while the opposing Nvidia and Meta are doubling down on hardware supply to the open-weights segment and on rolling out their own open models — suggesting that differences in business structure are manifesting as differences in stance.
Anthropic Signs $10 Billion Compute Deal with Volta
Anthropic has signed a contract to procure computing power from data centers managed by Volta Infra Holdings, an infrastructure upstart founded just months ago. According to people familiar with the matter, the deal is worth approximately $10 billion (on the order of 1.5 trillion yen). Volta is a cloud company backed by Nvidia, and Anthropic is scrambling to secure compute resources amid growing demand for Claude.
Taken together with Volta's previously reported fundraising moves, it is clear that AI developers are putting compute procurement at the very top of their priorities. The emergence of a mega-deal with a brand-new startup reflects the reality that the capacity of established major cloud providers alone can no longer cover demand.
CoreWeave Enters Asia for the First Time with 360MW in Indonesia
AI cloud provider CoreWeave is entering the Asian market for the first time. The company plans to build three data centers in Indonesia with a combined capacity of 360 megawatts, with operations expected to begin in 2028. The investment is expected to reach several billion dollars, though the exact amount has reportedly not been disclosed.
Asia is a region where demand for computing power is rising rapidly, and Western cloud providers have been announcing expansions one after another. CoreWeave's move can be positioned as part of that broader trend.
Dangerous macOS Vulnerability Went Unreported Because of AI Slop Reports
A symbolic security case was reported. A dangerous macOS security bug was left unreported for a long period, and one contributing factor was that Apple had been overwhelmed by massive volumes of AI-generated bogus bug reports (AI slop).
When vulnerability reporting channels are buried under auto-generated noise, genuine reports that should be reviewed first get lost in the flood. As AI-assisted report writing becomes widespread, this case shows the filtering cost on the receiving side surfacing as a new challenge. Similar issues have been pointed out before, including claims that Apple's bug bounty program has been rendered dysfunctional by AI slop.
AI Auto-Tracking Comes to Ukraine's Cheap Kamikaze Drones
Thanks to AI technology from a US company, low-cost kamikaze drones operated by Ukraine can now autonomously track targets. Beyond autonomous tracking, reports say expansion toward future swarm attacks is also in view.
The combat deployment of AI-based target recognition advances the efficiency of drone tactics, but it also brings back the ethical and legal questions of autonomous use of force without human judgment in the loop. Beyond technical progress, the debate over where to draw operational lines looks set to remain a focal point.
The 'Loyalty' Problem in the AI Talent War
In the talent war surrounding OpenAI, Google, Meta, and Anthropic, poaching battles are intensifying even as keeping people onboard becomes a challenge. A structure in which talent gathered with high compensation struggles to settle in has been pointed out, and some see the lack of 'loyalty' as becoming a bottleneck for running these organizations.
In an environment where movement among top-tier people has become routine, the question is how to retain talent through means other than compensation. As an issue that could even affect the sustainability of research and development, it looks like a structural challenge for the industry as a whole.