US DoD Signs Deals with 8 Companies for AI on Classified Networks, Anthropic Left Out
The US Department of Defense (DoD) has signed contracts with eight companies, including Nvidia, Microsoft, and AWS, to deploy AI on classified networks. The Pentagon has championed building "AI-first warfighting capability" and is pushing AI adoption across multiple vendors.
Notably, Anthropic is absent from the deal. The company was previously flagged as a security risk following a controversy over its terms of use. The DoD has emphasized a policy of vendor diversification, signaling an intent to avoid dependence on any single company.
The move shows that the line between AI companies and military use is being redrawn. Details on the terms each company agreed to have yet to be disclosed.
Uber Burns Through Its Annual AI Budget in 4 Months on Claude Code
According to Briefs, Uber burned through its entire 2026 AI budget in just four months through its use of Claude Code, Anthropic's AI coding tool. The news has become a major talking point in the developer community, racking up 70 points and 58 comments on Hacker News.
It's a case suggesting that the cost of AI coding tools has reached a level companies can no longer ignore. While productivity has improved, it has also underscored how difficult cost control can be.
Meanwhile, some voices say they "wouldn't pay more than $30 a month for AI coding tools," sparking debate over the gap in cost sensitivity between individual developers and enterprises.
Spotify Distinguishes Human Artists from AI with a "Verified" Badge
Spotify has begun rolling out a "Verified" badge to distinguish human artists from AI-generated content. Reported by the BBC, the initiative is positioned as a countermeasure against the surge of AI content in the music streaming industry.
With generative AI, anyone can now create high-quality music tracks, making it urgent for platforms to tell genuine artists apart from AI-generated works. Spotify's response could spread to other platforms as well.
Chinese AI Startups Dismantle Offshore Holding Structures and Move Toward Domestic Registration
According to The Decoder, Chinese AI startups such as Moonshot AI and StepFun have begun considering dismantling the holding company (VIE) structures they set up overseas. This comes after Beijing blocked Meta's acquisition of Manus, with the China Securities Regulatory Commission signaling a policy that companies hoping to go public should register domestically.
The move is seen as part of the Chinese government's tightening grip on the AI industry, and it could affect startups' fundraising and global expansion.
Bonsai: The First Commercially Practical 1-Bit LLM Arrives
"Bonsai," announced by PrismML, is billed as the first 1-bit (binary) LLM with real commercial viability. 1-bit quantization is a technique that enables extreme model compression, but significant accuracy degradation has been a longstanding challenge. Detailed evaluations are awaited to see how much quality Bonsai (8B parameters) retains.
Since lighter LLMs translate directly into lower inference costs, expectations are high for practical adoption on edge devices and embedded use cases.
This Week in The Batch: GPT-5.5 Delivers Strong Performance but Hallucination Issues Persist, Kimi K2.6 Tops the Open LLM Leaderboard
According to deeplearning.ai's weekly roundup "The Batch," GPT-5.5 showed improved performance, but hallucination problems remain. Kimi K2.6 also claimed the top spot on the open LLM leaderboard. Other topics included the impact of AI energy consumption on climate commitments and a comparison of strategic thinking between LLMs and humans.
---