Alibaba's AI office product Qianwen Office surpassed 30 million users in its first month, with enterprise users accounting for over half. The tool integrates document, spreadsheet, and presentation features with deep integration of Tongyi Qianwen's large language model for AI-assisted creation and editing.

Alibaba's Tongyi Qianwen office product, Qianwen Office, has reached 30 million users in its first month since launch, with over half being enterprise users. The product rolled out 120 version updates, averaging 4 iterations per day, and introduced over 1,000 new features. The international version is also live. The company open-sourced MyContext, an enterprise context infrastructure, which has gained over 3,000 stars on GitHub. Deep integration with DingTalk enables conversion of scattered business information into agent-accessible context. The dedicated office model Qwen3.8-Flash achieves 100% faster generation and 75% less token consumption. Enterprises like Chang'an Automobile have adopted the solution, reducing manual wire harness selection from two days to five minutes and vehicle document verification from over ten minutes to one to two minutes.

Alibaba Cloud's Qianwen Office has reached 30 million office users within its first month, with enterprise users accounting for over half. The platform completed 120 version updates, averaging 4 per day, and open-sourced its enterprise context infrastructure MyContext. Deep integration with DingTalk enables agents to query CRM and ERP systems. The dedicated office model Qwen3.8-Flash improves task speed by approximately 100% and reduces token consumption by 75%.

Alibaba introduced a glasses edition of Qoder, initially supporting Qianwen AI Glasses and Leqi AI Glasses. Users can assign tasks by voice, and the glasses capture error messages, whiteboard sketches, or screen info for the Agent to analyze. The interaction supports continuous dialogue without repeated wake words. The Qianwen version is in beta and expected to launch at the Yunqi Conference.

Bybit’s civil action tied to the North Korea-linked hack lays out a broad recovery framework built on RICO claims, John Doe defendants, emergency injunctions and sanctions coordination. But the article argues that a legal strategy, even a sophisticated one, does not mean stolen assets are already recoverable. At the time Bybit disclosed the lawsuit, it said about $48.4 million in stolen assets had been recovered and another roughly $30.5 million had been frozen at more than 28 exchanges and custodians, for a combined $78.9 million, or about 5.3% of the initial $1.5 billion loss. The piece separates tracing, control, proof of ownership and final return as four different stages, each with its own failure points. It also highlights limits created by asset commingling, cross-chain transfers, mixers, peer-to-peer transactions and third-party claims. One example involved Australian citizen Joseph F. Corrigan, who challenged an attempt to include a Nexo wallet holding about $39,000 in crypto in a preliminary injunction. The second half of the article then turns to lessons for Web3 firms: build joint response systems before an incident, prepare freeze-request templates and contact networks in advance, preserve chain data in a form courts can use, map real-world control points over assets, and tier recovery efforts across jurisdictions, sanctions exposure and expected value.

In the second phase of the "Qinglang" campaign, China's CAC has removed 5.61 million pieces of illegal information and shut down over 49,000 accounts. AI platforms including Doubao, Yuanbao, and Wenxin Yiyan (Ernie Bot) are now required to strictly limit harmful content generation. Social media and app stores are also strengthening content review mechanisms.

Ant Group CEO and Alipay chairman Han Xinyi used a recent China payment industry forum to sort global AI payment efforts into four camps: payment technology platforms led by Stripe, crypto firms represented by Circle and Coinbase, card networks led by Visa and Mastercard, and AI platforms represented by Google and OpenAI. The framing drew attention because Han explicitly described crypto as a medium for automated machine-to-machine payments, a notable formulation in a market where public messaging on cryptocurrencies has stayed cautious. The article ties that remark to Ant’s own push in AI payments. As of the end of May 2026, Alipay had processed more than 300 million AI payment transactions, with coverage across 95% of major general-purpose agent frameworks, including devices and platforms tied to Tongyi Qianwen, Rokid, Li Auto, Chery, Geely, Qoder and Coze. Han described the structure as one in which agents execute tasks while tokens carry value. The report also points to on-chain data showing where scale has emerged. Keyrock and Coinbase said AI agents completed about 176 million on-chain transactions from May 2025 to April 2026, with average ticket sizes of $0.31 to $0.48. Coinbase separately said its x402 protocol has handled more than 205 million transactions worth $53 million, with USDC accounting for 99% of on-chain agent commerce transactions.

Qianwen Office, an Alibaba-owned product, launched the newly released Qwen3.8-Flash model on the evening of Aug. 26, according to Techub News. The release was described as the first debut of the model on the platform, and it came alongside a standard mode introduced at the same time. The update centers on efficiency metrics disclosed in the brief: generation speed increased by 100%, while token consumption fell by 75%. The item cited Quantabit as the source attribution. No additional technical details, pricing information, rollout scope, or user availability terms were disclosed in the brief.