Prompt injections, the malicious commands attackers embed into content to entice large language models to follow them, have been attackers’ go-to tool for turning AI platforms against their users. A well-phrased command sneaked into an email or calendar invitation is often all it takes to cause the LLM to exfiltrate sensitive data or follow other harmful actions.

Now, defenders are embracing the prompt injection, too.

A strong, sharp effect

Researchers from Tracebit on Monday said they found that placing prompt injections alongside passwords, cryptographic keys, and other secrets stored on Amazon Web Services was often all that was needed to shut down attacks from AI hacking agents. The prompts direct the attacking LLM to perform an action forbidden by its guardrails, the safety barriers AI developers erect to prevent it from taking harmful actions. The LLM responds by shutting down.

Examples are a prompt that orders the LLM to provide steps for developing inhalable Anthrax spores, or, in the case of LLMs from Chinese developers, make references to the iconic Tank Man from the 1989 Tiananmen Square massacre. Once the LLM encounters these forbidden commands, it no longer follows its existing commands. The researchers have named the technique context bombing.

“Ultimately we’re triggering a refusal mechanism in the context,” Andy Smith, co-founder and CEO of Tracebit, said when explaining the name choice. “What we’re trying to capture is the fact that this does have a strong, sharp effect and one that can be difficult for the agents to come back from. Once they get that into their context they are going to keep refusing.”

Tracebit says initial testing suggests context bombing has great potential. They tested Opus 4.8, Gemini 3.1 Pro, GLM 5.2, DeepSeek 4 Pro, and Kimi 2.6 by giving them instructions to perform routine developer tasks that led the models to enumerate resources and stumble onto the planted strings. They ran the models inside a simulated AWS environment.

“Across five leading models and 152 attack runs, planting one of these strings in a decoy secret cut the rate at which agents seized full account admin from 57% to 5%, and complete compromise (where they also left themselves a persistent foothold) from 36% to 1%,” Monday’s post reported. “The most capable agent in our tests, Opus 4.8, went from achieving admin access in 93% of runs to failing every single time when confronted with a context bomb.”

Microsoft’s Satya Nadella says every firm using AI is paying for it twice, once in cash, and once in the secrets it hands over to make the thing useful. He calls it the Reverse Information Paradox. He also runs the company that helped build the trap.

Satya Nadella has a warning for everyone buying AI. You are paying for it twice. And the second payment is your crown jewels.

In a long essay on X that drew 10 million views, the Microsoft chief laid out an idea he calls the Reverse Information Paradox. It is sharp, a little wonky, and more than a little awkward coming from him.

Pay once in cash, once in secrets

The name is a riff on the Nobel economist Kenneth Arrow. Arrow’s original paradox was the seller’s problem. To sell information, you often have to reveal it, and once it is revealed, why would anyone pay?

Nadella flips it. In the AI age, he argues, the risk sits with the buyer. To make a model genuinely useful, you have to feed it your proprietary knowledge. The better you want it to work, the more you feed it.

So you pay in money, then again in something worth more: the know-how that makes your company yours. “The seller learns more and more about you as you use what you purchased”, he wrote, “while you learn very little about what the seller is learning in return.”

The leak you cannot see

The clever part is where he says the knowledge escapes. Not through some obvious breach, but through what he calls “exhaust”: the prompts you write, the tools your agents use, and above all the corrections you make when the model gets something wrong.

Every fix teaches the model. “It’s the kind of knowledge a competitor could never buy”, Nadella wrote, “and the kind that leaks almost imperceptibly: trace by trace, correction by correction, eval by eval.”

His verdict is blunt. If learning only flows one way, the money flows with it, toward whoever owns the AI, not whoever owns the knowledge.

The irony is doing a lot of work

Here is the catch. This is Microsoft talking.

Redmond poured billions into OpenAI and hosted ChatGPT on Azure. Its Copilot assistant is built to reach deep into a company’s email, files and chat. Back in 2024, roughly half of the data chiefs in one survey had paused or curbed Copilot over exactly this fear, as the Register noted.

To his credit, Nadella names his own side’s double standard. AI labs demand fair-use rights to train on the public web, then restrict customers from doing the same with model outputs. He is not wrong. He is also selling the fix.

Nadella’s answer, and his pitch

The solution, he says, is a hard “trust boundary” around a company’s data, evals and memory. Nothing crosses it, “not even the intelligence exhaust, without consent.” He borrows a line from Palantir’s Alex Karp about wanting to own the means of production.

His checklist runs to five points. Own your evals. Build learning environments inside your own tenant boundary. Keep the orchestration layer free of any single model. Then let it all compound. Microsoft, naturally, sells products that do each of these things.

Strip out the pitch and the core point still holds. This is the same executive who turned on the AI giants he helped build. The frontier labs are quietly amassing a fortune in other companies’ know-how. And the firms handing it over are, for now, doing it for free.

Of all the debates raging about the potential downsides of AI, there is one worry causing the most hand-wringing among AI enthusiasts in Silicon Valley. Their fear is that the giant AI labs that sell proprietary models are somehow acting like Trojan horses.

The concern is that, as startups and enterprises use AI models from labs like OpenAI and Anthropic, the labs gain ever-increasing access to those companies’ most sensitive business information. The model makers can then use that knowledge for themselves, potentially becoming competitors to their own customers. Those issuing such warnings range from VCs like Jason Calacanis to Palantir CEO Alex Karp.

Now, in a surprising blog post published on Monday, Microsoft CEO Satya Nadella has joined this crowd. Nadella warns that AI users (the “buyers” as he calls them) are paying twice. They knowingly spend for AI token usage but they also, obliviously, hand over valuable data in the process.

“You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!” he writes.

Most dangerously, enterprises are literally teaching the models about the nuances of their businesses, he argues.

“Models learn from ‘exhaust,’ the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how,” he writes.

This is “the kind of knowledge a competitor could never buy,” and yet enterprises are handing it over.

Nadella argues that if AI companies get to freely scrape the internet to train their models, it’s only fair that enterprises get to study — or “distill” — those models in return. “Distillation” is the practice of using a model’s own outputs to learn how it works and to train a new, often cheaper, model based on those insights. In February Anthropic accused Chinese open source models of sending millions of prompts to Claude as a way to improve their own models, and urged the U.S. government crack down on export controls.

Nadella’s point is that model makers can’t have it both ways. It’s hypocritical for them to freely train on the world’s data while restricting others from doing the same to their models.

“While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation,” the Microsoft CEO writes.

Nadella is particularly concerned when model makers “reserve the right to learn from customer usage and interaction data.”

Nadella’s solution is the kind of thing the CEO of a giant cloud provider would suggest. He wants companies to “retain ownership” of their data including prompts, feedback, etc. So he’s urging them to build their own “proprietary learning environments” on the cloud (where their data is likely already stored anyway and, conveniently, which could mean Microsoft’s cloud, Azure). He also wants companies to build in what he calls “orchestration layers” — essentially, a way to easily switch between AI models from different providers rather than being locked into one. Tools like AI “gateways” that let companies do exactly this, have become increasingly popular.

While Nadella never uses the words “open-source” as the method for retaining ownership, this is an obvious subtext. Yet, there’s another subtext.

Large companies, many of which still have some of their own data centers in addition to using the cloud, are already moving to open source models installed on their own premises (“on-prem,” in industry jargon). Idit Levine, founder and CEO of Solo.io — which makes networking and security software that helps enterprises manage AI systems — says she’s seeing exactly this shift play out with her own customers. After experimenting with proprietary model makers, they start asking themselves: “Can I take an open-source model and run it on-prem? It will do almost 90% of what the big one’s doing. It will cost way less,” she tells TechCrunch. “They understand that, and they can control it.”

Solo.io’s technology was selected last year as the tech powering the Linux Foundation’s Agentgateway project. Her company counts enterprises like T-Mobile, ADP and SAP as customers. She sees companies increasingly installing on-premise open source models and sees it as the next big wave in enterprise AI use.

She’s not alone. Vercel — best known as a platform for building and hosting websites, which has recently added AI model-switching tools — and OpenRouter, a company that helps developers route requests across different AI models — are both seeing a surge in traffic to open-source models. In fact, open models accounted for 29% of all traffic routed through Vercel’s gateway last month.

With the CEO of Microsoft, a company that has invested in both OpenAI and Anthropic, now openly urging enterprises to be wary of using proprietary models, we’ll bet this trend continues to grow. “In consuming intelligence, you are creating intelligence. And what you create should belong to you,” Nadella writes.

Tesla is building a wheelchair-accessible autonomous vehicle, a Tesla representative told lawmakers in Washington, DC, on Monday.

“We are in development for a purpose-built, wheelchair-accessible autonomous vehicle,” Tesla senior policy advisor India Herdman told members of the DC City Council on Monday, during a hearing focused on a controversial bill that could allow robotaxi services to operate in the District. “We know that paratransit can be very difficult, and people who are confined to wheelchairs permanently should still be able to move around freely, so that is an active product being built by Tesla in Texas,” she said.

Tesla didn’t respond to a request for comment. Herdman provided no further details about when a wheelchair-accessible product might be available. The electric automaker often takes several years to manufacture its announced products.

Tesla operates a small fleet of autonomous vehicles in the Texas cities of Austin, Dallas, and Houston and, as of this month, in Miami, Florida. (It also operates a service that uses human drivers in the San Francisco Bay Area.) The limited fleet uses Tesla Model Y, a compact SUV that is not wheelchair accessible.

The company has started to manufacture and test a purpose-built Cybercab, meant exclusively for autonomous driving and without steering wheels or pedals. These Cybercabs are not wheelchair accessible, though Tesla highlighted in an X post this month its accessibility features, including braille lettering on controls and wheelchair-height seating to allow for easier transfers.

Tesla and its CEO, Elon Musk, have hinted previously at a wheelchair-accessible autonomous vehicle. The company introduced an accessibility tab in its Robotaxi app last fall, though it directs users to other wheelchair-accessible ride providers in the area, rather than to Tesla’s own service. “We are working on accessible rides,” the app says. In response to an X user’s post last fall about Tesla working on accessible rides, Musk responded, “Absolutely.”

No US robotaxi company currently offers fleetwide driverless, wheelchair-accessible rides, including market leader Waymo. At the DC hearing on Monday, Waymo regional head of state and local policy Matt Walsh said, “To date, it’s my understanding that we haven’t been able to identify a platform that is fully wheelchair-accessible while also meeting the unique specifications to retrofit that vehicle with our technology.” He continued: “Now, I don’t want that to sound like a cop-out. We are trying to find that vehicle.”

Waymo has touted the accessibility features of its newest vehicle, the Zeekr-built Ojai, including its flat floor, low step-in height, and grab bars. But it is not wheelchair accessible. Michigan-based Ann Arbor autonomous-vehicle developer May Mobility offers rides in wheelchair-accessible vehicles in some of its markets, with a human operator on board to help deploy necessary ramps.

The Americans With Disabilities Act prohibits discrimination against people with disabilities in transportation services and requires reasonable modifications to provide equal access. Some but not all US cities require ride-hailing companies to provide wheelchair-accessible services. Many of those companies provide those rides through partnerships with specialized fleets made up of wheelchair-accessible vehicles.

In September 2025, the US Department of Justice sued Uber for “refusing to reasonably modify its policies, practices, or procedures where necessary to avoid discriminating against riders with disabilities.” The case is being litigated.

General Motors’ Cruise introduced a prototype wheelchair-accessible driverless taxi in 2023 and said it intended to roll out the vehicle in its self-driving car service in 2024. But following a collision with a pedestrian, Cruise all but halted national service in 2023. The next year, General Motors stopped funding its self-driving unit entirely.

Sonia Anand of McMaster University, Gina Ogilvie from the University of British Columbia and Vanessa Watts, also of McMaster University, explore the impact partial information can have on research in the health space.

Beyond its importance for scientific discovery, representation in health research directly influences healthcare planning, policy development and resource allocation.

Without research that reflects the diversity of the populations served, important health needs may go unrecognised, contributing to inequities in prevention, diagnosis and treatment. Without research, major health issues faced by a given community cannot be known, nor can effective interventions be developed if the populations most affected are underrepresented in the evidence base.

Research ethics boards (REBs) exist to ensure that research is conducted ethically and to protect participants from privacy breaches, coercion and exploitation. But in striving to achieve these goals, ethical board policies can sometimes produce an unintended consequence: over-protection that restricts the participation of people and communities historically excluded from research, such as racialised communities, Indigenous people and recent immigrants.

An ethics process must not become so preoccupied with the potential harm of participation that it overlooks the certain harm of exclusion.

REBs often struggle to strike the right balance between protection and inclusion. When the pendulum swings too far towards protection, it can result in the very injustice it’s meant to prevent. The consequences can be profound – the people said to be protected are not heard, counted or represented in the evidence that shapes policy, care and resources.

The Canadian Tri-Council Policy Statement (TCPS-2) governing Canada’s three largest public research funders is clear on this point. The statement asserts that “over-protectionist attitudes or practices of researchers or REBs, whether intentional or inadvertent, can exclude some members of society from participating in research”, and that such exclusion “may constitute a failure to treat them justly”.

It also states that researchers, institutions and REBs “must navigate between the dangers of imposing unfair burdens on particular participants, groups and communities, and overprotecting them”.

It’s not protection at all costs, but protection balanced with justice. Ethics recognise exclusion as harm, so caution must not erase participation.

Community-based research

This balance matters enormously in community-based research with vulnerable populations. One example from our work with newcomers to Canada in the Ontario city of Hamilton was designed precisely to overcome barriers to healthy active living, since obesity-related risks are shaped by the lived realities of migration, poverty, neighbourhood design and social exclusion.

The purpose of community-based research is to generate knowledge with communities, not merely about them, so that interventions can be relevant, usable and fair. But it’s the type of research that can be hampered by over-protective REBs. Examples include recruiting study participants in public spaces where newcomer families naturally gather, or requiring separate approval for each poster placed in the community.

The rationale is to prevent coercion of vulnerable populations. But exclusion is also an ethical risk. Regulations that make it so difficult to engage people that research cannot proceed don’t protect the community – they lock the community out of participation.

The harms of not doing research are rarely given equal weight in ethics deliberations. Careful attention is paid to the possible discomfort of participation, but non-participation also causes damage. When communities are excluded from research, there’s no ability to document the burden of disease on that community. We cannot demonstrate unmet need, measure inequity or build the evidence that directs resources, services and policy attention to the people who need them most.

In the absence of data, systems can always claim there is “not enough evidence” when the very structures of oversight have helped prevent that evidence from being gathered.

Respect, welfare and justice

The Canadian Tri-Council Policy Statement is rooted in three interdependent principles: respect for persons, concern for welfare and justice.

Concern for welfare requires researchers and REBs to protect participants from unnecessary risk and aim for the most favourable balance of risks and potential benefits. Justice requires that no segment of the population is unduly or overly burdened by research harms or denied the benefits of the knowledge generated from research.

Those principles cannot be selectively applied. If welfare is interpreted so narrowly that entire populations are denied the benefits of inclusion in research, then welfare has been severed from justice.

This is especially important for Indigenous Peoples. Research involving First Nations, Inuit and Métis people has too often been defined by non-Indigenous researchers, failed to reflect Indigenous world views and failed to benefit Indigenous Peoples or communities.

This is why Chapter 9 of the statement was developed; the policy emphasises respectful relationships, collaboration, engagement, representation in planning and decision-making, and attention to the specific situation of the community involved.

In other words, the answer to a history of exploitative research is not exclusion from research. It is better research: more respectful, participatory, accountable and more responsive to Indigenous priorities.

The historical relationship between research institutions and Indigenous communities has often been extractive. Unsanctioned medical experiments and the removal of material culture are examples that rightly justify stricter guidelines for Indigenous-focused research. And yet, there is a danger in overstating harm through the blanket framing of Indigenous Peoples as inherently “vulnerable” when REBs assess proposed studies.

Many Indigenous communities and organisations maintain their own research ethics and governance infrastructures, including data governance protocols that reflect principles of Indigenous data sovereignty. Indigenous researchers themselves must abide by these protocols. Indigenous data sovereignty emphasises community and nation-based authority over when, how and by whom data are accessed, shared or withheld.

These infrastructures have expanded in response to the increasing demand for Indigenous research following the release of the Truth and Reconciliation Commission’s Calls to Action. Universities now have more Indigenous researchers, more interest in Indigenous perspectives, more expectations of Indigenous participation and consultation, more calls for collaboration and co-creation, more frameworks attentive to ethics, consent and privacy.

Vulnerability, where it exists, should not be understated – real risks remain and must be mitigated. However, REB ethical vigilance should not become automatic exclusion.

Balancing protection and exclusion

Another example of over-protection is the exclusion of pregnant women from clinical research. There may be good reasons to exclude pregnant women from some studies, such as trials of medication that could potentially harm a fetus. But in many other cases, exclusions leave clinicians and patients without evidence for the people who actually need care.

Excluding a group to avoid risk may itself create long-term injustice by making the evidence base less applicable to them.

What should change is not the abandonment of an ethics review, but a more careful balance between protection and the injustice of exclusion. REBs and administrators should ask, at the outset, what harms may follow if this research is not done, is delayed or is made infeasible? Who loses if recruitment cannot happen in the spaces where people are? Whose voice disappears when the default answer is no?

The people most often missing from decision-making are the newcomer parent, the Indigenous community member, the racialised participant and the pregnant woman with no spare time to navigate institutional obstacles. These are not the people we should make hardest to reach. These are the people we should work hardest to include.

We must consider other ethical failures in addition to exploitation: the injustice of being excluded, including failure to measure the burden of disease affecting specific communities, and ensuring the safety and efficacy of treatments in that community.

Unless exclusion itself is recognised as a form of ethical injury, REBs, institutions and funders risk sidelining the people research is most needed to serve.

Engaging with communities of interest can help strike the right balance, whether through representation on REBs, representation on research teams or opportunities for community representatives to speak directly to the urgency of the research. Ethical oversight must do both: protect participants from harm and protect communities from being left out.

By Sonia Anand, Gina Ogilvie and Vanessa Watts

Sonia Anand is an associate vice-president for global health at McMaster University, a professor of medicine and epidemiology and a vascular medicine specialist at Hamilton Health Sciences in Canada. She holds the Canada research chair in ethnic diversity and cardiovascular disease, and is the Heart and Stroke Foundation of Ontario/Michael G DeGroote chair in population health research. Her present research focuses on the environmental and genetic determinants of vascular disease in populations of varying ancestral origin and women and cardiovascular disease.

Gina Ogilvie is a Tier 1 Canada research chair in the global control of HPV-related diseases and prevention, and a professor at the University of British Columbia in the School of Population and Public Health. She is also the associate director of the Women’s Health Research Institute at BC Women’s Hospital and Health Centre. She is currently principal investigator for more than $10m in research grants and has received funding from NIH, PHAC, CIHR, the Michael Smith Foundation for Health Research, Canadian Foundation for Innovation and private foundations including the BC Women’s Hospital Foundation.

Vanessa Watts is an associate professor of indigenous studies and sociology at McMaster University. She teaches about contemporary indigenous issues, residential schools, indigenous sovereignty, indigenous ways of knowing and methodologies, and indigenous ontologies. In 2018, she was awarded a SSHRC Insight Development Grant for her project, ‘An Indigenist Sociology of Knowledge: Indigenous social lives in Indigenous studies, sociology and political science (1895 and beyond)’. She is also a research fellow at the Yellowhead Institute at Ryerson University.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

If you’ve ever been hooked up to a wearable machine at a doctor’s office, then you’re familiar with the electrodes that are attached to your body to monitor its electrical signals. The problem with these prefabricated metal-based or hydrogel electrodes is that they don’t always stay in place during movement, for long periods or on sweaty or hairy skin. 

Penn State University engineers aim to change this with paint-on tattoos that use conductive ink to power sensors for wearable devices such as EEGs, ECGs and EMGs that track brain, heart and muscle activity, respectively. 

As reported in a paper published Monday in the Proceedings of the National Academy of Sciences, the engineering team filed a provisional patent for this ink, a water-based solution mixed with polymers and acidic additives that starts out transparent with a glue-like consistency. It can be pigmented with food dye to create different colors for a cute fox or shark that opens its mouth when you open your hand, and it dries on the skin in under 10 minutes. It can later be reapplied or washed off. 

Thanks to its customizable, fun nature, these paintable electrodes could be especially beneficial for children who may be more likely to wear a medical device if it’s powered by a temporary tattoo of their favorite character or animal. 

How the ink powers wearable devices

To connect the ink to sensors, there’s a porous silver textile with connective electrodes. Before your painted-on tattoo dries, the textile is placed on the design so that it will stick to the skin. Then, the textile is connected to a port on a wearable monitoring device. The latter is taped to the skin beneath clothing.

Electrical signals collected by the ink are sent through the textile to the monitoring device, which then transmits the data to a computer via Bluetooth. 

“The big idea behind this is that in the future, you could potentially have a more expensive sensing module that remains separate from the system, but the electrodes themselves can be disposable. A single bottle of ink could provide enough material to paint multiple electrodes over the course of several days or a week,” said Larry Cheng, the paper’s corresponding author and a James L. Henderson Jr. memorial professor of engineering science and mechanics at Penn State, in a press release.

The hope is that these temporary electrode tattoos can help spot heart attacks early, read brain waves or power robotic prosthetics.

Plus, since it’s painted directly onto the skin, the ink is more durable and accurate than sensors attached to the skin, as there may be an air gap between them and the skin. As for the silver textile, because it’s porous, its connective electrodes can expand to over 150% their original size. This means sweat can pass through the textile without negatively impacting adhesion, accuracy or comfort. 

What the team found during testing

When experimenting with the ink, the Penn State team found that the painted electrodes could monitor ECG signals for up to 12 hours. They also stayed on during exercise. When applied to a team member’s forearm, the electrodes successfully tracked muscle signals using an EMG device, enabling remote control of a robotic hand. 

Since the electrodes can be washed off and reapplied, 12 hours isn’t the limit for their use. 

Eventually, the painted electrodes could even power sensors that track cortisol or glucose. The team is also looking toward commercial use for doctors, such as pediatricians. Or, to create “smart plants” that provide information on chemical exposure in their environment and its impact on plant health. 

Just like it’s become normal to see people wearing health tracking smartwatches and smart rings, perhaps one day no one will bat an eye when they see someone leaving a doctor’s office with a temporary tattoo. 

HiFi Rose has officially launched the RW800 ROSE AIR Lite, a compact streaming expansion module designed to bring newer network audio features to existing Hi-Fi systems without forcing owners to replace the hardware they already own.

That sounds simple enough. It is not.

The RW800 is both a standalone network streamer and an expansion module for compatible HiFi Rose components. In standalone mode, it can connect to an amplifier, active loudspeakers, or an external DAC. In expansion mode, it connects to select HiFi Rose products and adds modern streaming support that some owners have been asking about for a long time.

HiFi Rose announced the RW800 through its official community site on July 7, confirming that pre-orders are available through authorized dealers and that a launch promotion for existing HiFi Rose customers runs through September 30, 2026. Regional pricing and promotion details will depend on local dealers, so U.S. buyers should confirm availability and final pricing with an authorized HiFi Rose retailer.

Related Reviews:

Why the RW800 Matters

HiFi Rose has built its reputation on large touchscreen streamers, attractive industrial design, and a software platform that combines music, video, internet radio, local playback, CD ripping, and streaming services. The problem is that streaming services do not stand still. TIDAL Connect, Qobuz Connect, Spotify Connect, AirPlay 2, Google Cast, and Roon Ready support are no longer fringe features. For many listeners, they are the way they actually use a streamer.

That has created tension inside the Rose ecosystem.

HiFi Rose owners have been asking about TIDAL Connect and broader Connect-style support for years. In April, a HiFi Rose representative said the company encountered technical difficulties trying to add TIDAL Connect to existing hardware and developed the RW800 to support TIDAL Connect and other functions that could not be added through firmware updates alone.

That makes the RW800 more than another small streamer. It is HiFi Rose’s answer to a real platform problem.

For current Rose owners, the RW800 may be the most practical way to add features that competing streamers already include natively. That is both useful and slightly awkward. When someone spends serious money on a network player, they expect the software side to age gracefully. The RW800 suggests that HiFi Rose knows some customers need a bridge to get there.

What the RW800 Does

The RW800 supports TIDAL Connect, Qobuz Connect, Spotify Connect, AirPlay 2, Google Cast, Roon Ready, Bluetooth 5.2, dual-band Wi-Fi, Ethernet, and USB audio. HiFi Rose also lists high-resolution playback support up to PCM 32-bit/384kHz and DSD128.

The key feature is its two-mode design.

In Expansion Mode, the RW800 works with compatible HiFi Rose components, adding newer streaming capabilities to the existing Rose system. HiFi Rose has specifically discussed compatibility with products including the RS130, RS150, RS451, RS151, and RS201. Dealer materials also list the RS250A among compatible products.

In Standalone Mode, the RW800 acts as a compact network streamer for non-Rose systems. That means it can be used with an integrated amplifier, powered speakers, or an external DAC. For someone with an older but still excellent audio system, that may be the cleaner use case: keep the amplifier and speakers, add the streaming layer, and avoid buying a larger touchscreen streamer just to access modern services.

RW800 Core Features

The RW800 is small enough to hide in a rack or place on a desktop, but the feature set is broader than the footprint suggests.

• Product type: Streaming expansion module and standalone network streamer
• Operating modes: Expansion Mode for compatible HiFi Rose components and Standalone Mode for other audio systems
• Streaming support: TIDAL Connect, Qobuz Connect, Spotify Connect, AirPlay 2, Google Cast, Roon Ready, DLNA/UPnP, and Bluetooth
• Playback support: PCM up to 32-bit/384kHz and DSD128
• Wireless and network connectivity: Bluetooth 5.2, dual-band Wi-Fi, and Ethernet
• Digital connectivity: Optical input and optical output, with USB audio used for Rose system integration
• Analog output: Stereo RCA output for connection to an amplifier or powered speakers
• Control: ROSE AIR app for iOS and Android

Who Is It For?

The RW800 makes the most sense for three groups of listeners.

The first group is existing HiFi Rose owners who like their hardware but want newer app-based streaming support. If you own a Rose streamer and have been waiting for TIDAL Connect, Qobuz Connect, Spotify Connect, AirPlay 2, or Google Cast support to become easier to use, the RW800 is aimed directly at you.

The second group is anyone with an older amplifier, active loudspeakers, or external DAC who wants a compact streaming front end without buying a full-size network player. In that context, the RW800 is less of a Rose accessory and more of a small streamer that happens to come from a brand known for premium network audio products.

The third group is the listener who wants Roon Ready, Connect-style streaming, and Bluetooth in a small box but does not need HDMI ARC, a large touchscreen, phono input, room correction, or preamp features.

That last point matters because the RW800 is not trying to be a WiiM Ultra, Bluesound NODE, Cambridge Audio CXN100, or Eversolo DMP-A6. It is more focused than that. Whether that focus is a strength or a limitation will depend on the system.

What Makes It Unique

Most affordable streamers are designed as standalone boxes. The WiiM Ultra, Bluesound NODE, Cambridge Audio MXN10, Volumio Primo V3, and Eversolo DMP-A6 all expect to be the streamer in the system.

The RW800 is different because it also works as an ecosystem extension. It can serve as a compact standalone streamer, but its more unusual role is as an add-on module for existing HiFi Rose owners. That is not common, and it gives the RW800 a very specific reason to exist.

There is a fair criticism here. Should owners of expensive streamers need an external module for features like TIDAL Connect and Spotify Connect? Probably not. But there is also a practical answer: if the main Rose hardware still performs well and the missing features cannot be added cleanly through software, a compact add-on is better than telling customers to replace an entire component..

Competitors?

The WiiM Ultra is the obvious value rival for anyone starting from scratch. It offers a touchscreen, broad streaming support, HDMI ARC, phono input, subwoofer output, room correction, and preamp features at a very aggressive price. If you do not need Rose integration, the WiiM Ultra remains one of the hardest products in the category to ignore.

The Bluesound NODE is the more established multiroom alternative. BluOS remains one of the strongest streaming platforms in consumer audio, and the current NODE is a better fit for users who want HDMI eARC, app maturity, broad service support, and a larger ecosystem of wireless speakers and components. eCoustics recently reviewed the Bluesound NODE as a $750 streamer competing directly against WiiM and Cambridge.

The Volumio Primo V3 is another relevant comparison because it supports TIDAL Connect, Qobuz Connect, Spotify, Roon Ready operation, and Volumio Premium in a more conventional standalone streamer. eCoustics covered the Primo V3 as a streamer focused heavily on software flexibility and app-based control.

The Ferrum BROEN sits higher up the food chain as a network streaming transport for external DAC users. It is powered by Volumio and supports Spotify Connect, TIDAL Connect, Qobuz Connect, AirPlay via Shairport Sync, UPnP/DLNA, Bluetooth playback, USB storage, NAS playback, and CD playback and ripping. It is not really the same buyer as the RW800, but it belongs in the broader conversation about external streaming bridges and transports.

The Eversolo DMP-A6 Gen 2 and related Eversolo models are the larger-screen alternatives for buyers who want a more complete standalone component with local storage options, touchscreen operation, and a fuller hardware interface. The RW800 is smaller and more focused. Eversolo is more of a full-featured front end.

The Bottom Line

Streaming compatibility now matters as much as the DAC, chassis, power supply, and display. For many listeners, the best streamer is not the one with the most impressive spec sheet. It is the one that works cleanly with the services they already use.

The RW800 gives HiFi Rose owners a path to newer streaming features without replacing their existing hardware, and it gives non-Rose users a compact way to add modern network playback to an older system. It also raises an uncomfortable but necessary question about how long premium streamers should remain current in a software-driven audio world.

HiFi Rose deserves credit for offering a solution. Whether the RW800 feels like a clever upgrade or a workaround will depend on the system, the owner, and how much patience they have left for streamer software drama. In 2026, that patience is not exactly in surplus.

For more information: hifirose.com/product/rw800

Related Reading: