Qwen's Personal Agent wants your health data and your portfolio

Share
Qwen's Personal Agent wants your health data and your portfolio

Alibaba spent its Hangzhou conference week talking about chips, gigawatts and ten-trillion-parameter roadmaps. The consumer half of the story is quieter and more consequential: what the Qwen app does next with your data.

Alibaba is rebuilding its Qwen assistant as a "Personal Agent" that runs on authorised health and exercise data and, for some users, linked brokerage holdings — and it brought efficiency numbers to argue the plumbing is ready. Qwen product lead Zheng Sishou said at the Yunqi conference that the app is being rebuilt around the Qwen 3.8 series' agentic abilities for its 300 million users. In health, it builds a long-running personal context from workout history and connected devices; the demo Alibaba chose is deliberately mundane — set a sub-two-hour half marathon, get daily training and diet adjustments from logged runs and a linked glucose monitor, and after you hit the pace, get told to push for 1:40. In finance, Qwen blends live market data, institutional research and a user's own positions into personalised analysis, and wraps financial firms' research output as Skills. The company says dedicated Harness tuning lifted response quality on finance tasks by 20%, and that dynamically adjusting the Harness saves each user 40% of their token spend.

That second number is the one worth keeping, because a personal agent that reads continuous data every day is a cost problem before it is a product. Storing months of context per user and re-reading it on every question is exactly where consumer assistants have quietly failed to scale, and Alibaba is claiming a 40% cut by deciding at runtime how much context to assemble. It is also the clearest statement yet of the difference between a chatbot and an agent: one answers, the other holds state about you over months.

The ecosystem is being widened at the same time. Qwen's open platform has moved from local services and logistics into jobs, sports and gaming — more than 20 fields, with over 1,000 partners applying — and the company says more than 60% of users who bring it finance, health or life questions go on to use the work assistant. We covered the enterprise end of this push when Ant Group rolls Qianwen Office out to every employee, and on Tuesday Alibaba added that Qwen Office will plug into Salesforce on Alibaba Cloud, so sales staff can query opportunities and maintain customer records from inside the assistant. Health records and brokerage positions are the two categories where "convenient" and "surveillance" are hardest to tell apart, and Alibaba's case rests entirely on authorisation being explicit, revocable and per-dataset.


The US embassy in Canberra has filed a rare submission calling Australia's draft digital duty of care laws "censorship of protected speech." Under the draft, platforms risk hefty fines unless they let users switch off the algorithms that decide what their feeds show. Washington asked Australia to clarify how "harm" and "risks" would be determined, argued that mandated platform design features "may affect what users see, say, and hear not just in or from Australia, but globally," and warned the rules could damage Australia's reputation as a jurisdiction that enables innovation.

Prime Minister Anthony Albanese, in New York for the UN General Assembly, rejected the framing: "It's not about giving government control. It's about giving people back control over what they receive on their devices." The submission is the same instrument Washington used on allied media policy earlier this month, when The US calls the UK's plan to rank "trustworthy" news censorship — a formal filing aimed at a domestic law it cannot vote on. Australia's opt-out is the stronger of the two designs on paper, because it gives users a switch rather than a regulator a ranking, which makes the "censorship" label harder to land.

What to watch: whether the opt-out survives into the bill Australia actually introduces, and whether Qwen's health-data integrations ever ship outside mainland China — they will not, which is the quiet advantage Western assistants have.

If an assistant holds your workouts and your portfolio, what would make you trust it enough to switch it on? Tell us in the comments.

Read more

Brain organoids classified Braille at 83% — in a trio, not alone

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base. The run itself is the number worth keeping: 16,000 Nvidia GB200 GPUs for 38 days across August and September, at 21% average model

Amazon's new Bee devices listen all day — for $11.99 a month

Amazon's new Bee devices listen all day — for $11.99 a month

Amazon turned its always-listening Bee line into a two-device family on Thursday, and NVIDIA opened a 2.8-trillion-token Indic corpus recipe. The connecting thread: both are bets that the next AI money is in memory — yours, or your model's. Amazon put two new always-listening devices on sale, and both now lead to a subscription. Bee Frontier, the next-generation wristband that also clips to clothing, ships now at $74.99 in Onyx, Lavender and Sage — up from the $50 predecessor — with seven-day

Perplexity open-sources its agent sandbox after 118 failed escapes

Perplexity open-sources its agent sandbox after 118 failed escapes

Perplexity handed its agent containment layer to the open-source world, and Washington's AI task force stopped reading briefings and started working. Perplexity open-sourced SPACE on Thursday — the sandbox it built to keep its own AI agents from getting out — and says nine frontier models failed to break out of it across 118 attempts. Each agent runs inside a small virtual machine with its own operating system, with credentials and network access managed outside the sandbox so an agent can re