The corporate walls go up around AI coding assistants
The day's dominant thread was a wave of companies restricting which AI coding tools their engineers may touch. Chinese tech giant Alibaba told employees to stop using Anthropic's Claude Code, adding it to a list of high-risk software amid allegations that the tool contains a hidden "China-detection" backdoor. Multiple reports — from Business Today, Market Chatter and others — converged on the same story, with Alibaba reportedly steering staff toward its own in-house coding assistant, Qoder, instead.
Alibaba wasn't alone. Meta Platforms is reportedly reining in its engineers too, restricting them from using both Anthropic's Claude and OpenAI's Codex. Whatever the specific motives — security concerns, competitive worries, or a preference for home-grown tools — the message is consistent: some of the world's largest tech employers are now treating rivals' AI assistants as something to keep off the corporate network.
Meanwhile, Claude moves deeper into the mainstream
Cutting against that grain, Anthropic's Claude models became publicly available inside Microsoft Foundry, Microsoft's platform for building and deploying AI applications. The models are hosted on Microsoft Azure and, notably, run on Nvidia's newest chips — a reminder that even as some firms wall Claude out, the biggest cloud platforms are racing to fold it in.
Beijing tightens the rules, and a new rival emerges
China's regulatory hand was visible elsewhere. The country's most prominent chatbots are set to shed their human-like "AI personas" as Beijing tightens its rules on artificial intelligence, per Nikkei Asia — a small but telling sign of how differently AI is being shaped there. At the same time, a Chinese startup called Z.ai is increasingly being cast as a challenger to America's two best-known AI companies, OpenAI and Anthropic.
Regulators on both sides of the Atlantic circle AI
Governments are catching up. Britain's top financial regulator said the country should consider whether large language models — the systems behind ChatGPT, Claude and Gemini — need to fall under financial regulation. In the US, the fight over who governs AI is coming to a head: the Trump administration unveiled a national AI policy framework designed to limit state power over the technology, aiming to rein in a sprawling patchwork of some 1,200 state-level AI bills.
Ransomware that thinks for itself
On the security front, researchers documented what may be a genuine turning point: a strain dubbed JadePuffer, described as the first ransomware that behaves less like a rigid script and more like an autonomous operator that thinks on its feet. If the analysis holds, it signals a worrying new chapter in AI-enabled cybercrime.
Quiet progress in quantum and science
Two quieter stories rounded out the day. IQM Quantum Computers, a self-described global leader in full-stack superconducting systems, acquired selected assets of German firm Quantistry GmbH — a move to better connect quantum chips with industry labs. And a grounded reminder from The Transmitter: AI is genuinely reshaping how science gets done, being folded into the everyday work of research to strengthen scientific processes — just not in the flashy, hype-driven ways headlines often promise.
Taken together, it was a day of contradictions: AI assistants banned in one boardroom and embraced in another, regulators reaching for the rulebook, and the technology quietly maturing in labs even as its darker uses grow more capable.