Plaud wants its new AI earbuds to record and process workplace conversations without relying on a nearby smartphone.

The company’s new Plaud One Explorer Edition pairs wireless earbuds with a charging case that includes an eSIM and 4G LTE connectivity, allowing recordings to be uploaded to Plaud’s AI services without relying on a nearby smartphone or Wi-Fi.

Plaud says the case can capture in-person conversations, while the earbuds can record phone calls and online meetings. Plaud Intelligence then transcribes and summarizes the audio. From there, Plaud Agent is designed to use that context across connected services such as Gmail, Google Calendar, Notion, and Slack, turning conversations into follow-up emails and documents.

How Plaud’s AI earbuds work

Plaud One combines two familiar pieces of hardware — a pair of wireless earbuds and a charging case — with microphones specifically designed to capture speech.

According to Plaud, the earbuds have a recording range of up to 2 meters and can record in-person conversations for up to 6 hours.

Plaud says the case has four microphones, a pickup range of up to 5 meters, and enough battery life for as much as 25 hours of continuous recording, allowing it to capture in-person discussions without the earbuds.

The more unusual part is the case’s eSIM and 4G LTE connectivity, which lets the system send captured audio to Plaud’s cloud services without relying on a nearby smartphone or Wi-Fi connection. Once uploaded, Plaud’s AI can transcribe and summarize the recording, turning a conversation into structured information that its AI features can later use.

The important distinction is that Plaud isn’t treating the earbuds primarily as an audio accessory, but rather a hands-free input device for its AI system. That effectively makes it an AI-first device.

Agentic AI is Plaud’s biggest ambition

Plaud’s bigger bet is turning those recordings into actions, not just summaries.

Plaud says its Agent features will eventually use recorded context to draft emails, create documents and presentations, and work across services such as Gmail, Google Calendar, Notion, and Slack. Broader agent functions and integrations based on MCP are expected through later software updates.

AI earbuds move beyond audio

Plaud is not alone in putting AI at the center of earbuds.

Timekettle’s W4 Pro AI Interpreter Earbuds, for instance, use AI for real-time two-way translation, phone-call translation, and online-meeting translation.

Apple appears to be pushing the idea further, with reports pointing to camera-equipped AirPods that could feed visual information to Visual Intelligence, letting Siri understand and respond to objects in the wearer’s surroundings.

Taken together, these products suggest earbuds are expanding beyond AI-assisted audio into devices that can capture, interpret, and act on information from a wearer’s surroundings.

More must-read AI coverage

How to get the Plaud One

The Plaud One Explorer Edition is available for preorder in selected markets for $249.99, with shipping expected to begin in the fourth quarter of 2026.

Plaud’s free Starter plan includes 300 transcription minutes per month. Customers who need more can upgrade to a paid plan; TechCrunch lists the Pro plan at the equivalent of $8.33 per month when billed annually. Pricing and allowances may vary by region.

The privacy implications of Plaud One

Plaud’s biggest challenge may not be getting people to see the value in AI earbuds, but getting everyone around the wearer to accept them as well.

Plaud told TechRadar that recording must be started manually and that users are responsible for following applicable consent laws. Higher-impact agent actions, such as sending emails or editing files, will require user approval.

But the privacy concern is bigger than Plaud.

Meta has faced frequent scrutiny over recording through its AI glasses, while reported camera-equipped AirPods are already drawing similar privacy questions before Apple has even announced them.

And that raises a bigger question for AI wearables: if people and regulators are increasingly uncomfortable with devices that can capture their surroundings, why do companies keep putting them into everyday objects? The answer may be that the convenience is strong enough to sustain demand, but the industry still has to convince everyone around the user that these devices are trustworthy.

Plaud plans to release the Plaud One in limited quantities, according to TechCrunch. Prospective buyers should remember that the device has not yet shipped and that several advertised agent capabilities depend on future software updates. For businesses, its value will depend not only on how accurately it records meetings, but also on whether Plaud can provide sufficient privacy controls, predictable AI costs, and reliable integrations.

Read more: Singapore’s growing use of AI glasses shows how wearable recording devices are forcing workplaces and regulators to reconsider existing privacy rules.

With on-site service dispatches often adding hundreds of dollars in labour and travel costs, managed connectivity and technology solutions provider BCN has launched a managed service designed to detect connectivity failures automatically and then remotely recover network equipment and restore service before an on-site visit becomes necessary.

With the launch of the Remote Power Monitoring, Management and Cycling (RPM) system, BCN cited research from 2024 showing that major IT outages can cost organisations more than $300,000 per hour.

In addition, it noted that despite advances in cloud computing, artificial intelligence (AI), software-defined wide-area networks (SD-WAN) and edge infrastructure, one of the most common causes of network downtime still ends the same way: someone has to manually reboot network equipment. BCN stressed that the increased adoption of cloud applications, AI, unified communications, internet of things (IoT) and edge computing has made reliable connectivity more important than ever.

As a solution, BCN RPM is claimed to be able to help organisations improve network resilience through intelligent automation while extending BCN’s managed connectivity experience. It is designed to help organisations reduce downtime by automating recovery from common equipment issues and minimising unnecessary truck rolls. It builds on the company’s core managed connectivity strategy by adding intelligent power monitoring and recovery.

And while it accepts that remote power management technology has existed for years, BCN assured that it could create value by integrating technology into a managed service that would improve visibility, uptime and the overall customer experience.

That is to say, instead of waiting for someone to drive to a location simply to unplug and reconnect a router or modem, RPM continuously monitors network health and is able to initiate recovery automatically based on configurable network conditions. The service is also said to be able to allow businesses to remotely reboot equipment on demand or schedule preventative maintenance across multiple locations through a secure management portal.

Other key benefits include automatic recovery when internet connectivity is lost; remote reboot of routers, modems, firewalls, switches and access points; scheduled maintenance reboots during off-hours; and independent control of multiple powered devices. As well as secure remote management through web and mobile applications, the system is attributed with a reduction in support tickets and onsite service dispatches, and higher network availability across distributed locations.

“Reliable connectivity has become essential to every modern business, and our responsibility does not stop at delivering the network. It extends to helping customers maintain the infrastructure that supports it,” observed BCN chairman and CEO Richard Boudria.

“RPM represents another step in BCN’s vision of providing a complete managed connectivity experience, combining intelligent monitoring, proactive management and operational expertise to reduce disruptions and keep businesses connected when it matters most.”

BCN also regards RPM as extending its managed connectivity portfolio by integrating enterprise-grade remote power monitoring into a fully managed service. The company’s engineers are available to design, configure, monitor and support the solution throughout its lifecycle, giving customers a single accountable partner for network resilience.

It also adds to a portfolio of managed connectivity, SD-WAN, cyber security, cloud communications, network-as-a-service (NaaS) and global connectivity solutions.

Qualcomm Technologies has revealed an intent to establish a long-term investment initiative designed to accelerate innovation across Japan’s tech ecosystem.

Qualcomm Technologies’ investment framework is designed to support the entire robotics innovation lifecycle, from early stage technology development through to commercial deployment and global expansion. The firm aims to work alongside industry, academia and startups to help position Japan at the forefront of the intelligent robotics era.

The initiative combines four strategic pillars spanning innovation, ecosystem development, commercialisation and deployment to help accelerate the development, deployment and global scaling of next-generation robotics services. It is also aligned with Japan’s national policy priorities to address labour shortages, strengthen industrial competitiveness and accelerate the adoption of artificial intelligence (AI) and robotics across key industries.

In what is described as “a cornerstone” of this effort, the technology firm will establish the Qualcomm Japan Robotics Centre, a hub for collaborative research, development, ecosystem collaboration and engagement, workforce enablement, and robotics innovation. Among its aims will be advancing innovation, strengthening industrial competitiveness, and accelerating the development and deployment of next-generation physical AI and robotics offerings.

An ecosystem scaling programme is designed to accelerate ecosystem growth by providing startups, developers, researchers and industry with access to Qualcomm Technologies’ products, technical expertise, enablement resources and what could be collaborative innovation opportunities.

A global go-to-market accelerator is designed to help Japanese robotics innovators scale beyond domestic markets by connecting them with global customers, partners and deployment opportunities.

Finally, a lighthouse initiative is designed to support the deployment of flagship physical AI initiatives that demonstrate the transformative potential of next-generation robotics and intelligent automation across Japan.

Organisations supporting the initiative include AI Robot Association, Amatama, APTO, Fanuc Corporation, Fixstars Corporation, Fujitsu, Kawada Technologies, Kawasaki Heavy Industries, KDDI Corporation, Minebea Mistumi, NEC Corporation, NTT Docomo, Panasonic Industry, Preferred Networks, Robotics, Sony Corporation and Toyota Motor Corporation. In addition, academic institutions such as The University of Tokyo’s Fukui Lab will help support research, talent development and next-generation robotics.

Nakul Duggal, executive vice-president and group general manager of automotive, industrial and embedded internet of things and robotics at Qualcomm Technologies, stressed how long Japan has been a global leader in robotics and industrial innovation.

“Through this investment initiative and the establishment of the Qualcomm Japan Robotics Centre, we are intending to create a foundation for deeper collaboration across industry, academia and startups,” he said. “Together, we aim to help accelerate the development, deployment and global scale-up of next-generation physical AI and intelligent robotics solutions.”

Amatama CEO Yuichi Kachiyama added: “Physical AI will not scale on cloud compute alone. Intelligence has to be arranged hierarchically inside the machine: at the periphery where the sensors sit, at the intermediate layer where coordination happens, and at the centre where intent is formed.

“Through the Qualcomm AI Program for Innovators, we have already completed a proof of concept for nHOS on Qualcomm Dragonwing AI 100 Ultra and confirmed its feasibility. We look forward to extending that work at the Qualcomm Japan Robotics Centre, and to proving out from Japan an open foundation for robotics that the world can build on.”

AI benchmarks are supposed to reveal what models can do, but Google DeepMind is now putting the tests behind a cryptographic wall to make sure the models have not seen the answers first.

Google DeepMind said Thursday that it has piloted what it describes as the first double-blind evaluation of a proprietary frontier AI model, using a cryptographically protected environment to keep both the model and evaluation prompts hidden from each side.

The project involved the Singapore AI Safety Institute, OpenMined, AVERI and MLCommons. AVERI evaluated Gemini 2.5 Flash Lite using reserved prompts from MLCommons’ AILuminate safety benchmark, covering cyberattacks, chemical and biological hazards, hate speech, self-harm and violent-crime elicitation. Singapore AISI separately tested the model using confidential prompts focused on harmful content in Singapore’s context.

The setup addresses a growing problem in AI testing: benchmark contamination. If a model or its developer has access to test questions before an evaluation, a strong score may reflect familiarity with the benchmark rather than the model’s underlying ability.

Google said traditional protections such as zero-logging policies and contractual restrictions have helped keep evaluation prompts confidential, but cryptographic safeguards can add another layer of protection.

Neither side gets to peek

The system uses Google Cloud’s Confidential Computing technology to place the model and evaluation data inside a protected environment.

The evaluator cannot access Google’s model weights, while Google cannot access the evaluator’s test prompts. The pilot ran on a Google Cloud A3 Confidential VM using Intel TDX host-memory encryption and an NVIDIA H100 Confidential GPU. Hardware encryption and remote attestation were used to keep the benchmark prompts and model weights isolated while verifying the software environment.

The approach is designed to reduce a long-standing trade-off in external AI testing: evaluators previously had to either provide sensitive test material to model developers or ask companies to expose proprietary model weights. Google DeepMind said the approach could be particularly useful for sensitive evaluations involving cybersecurity and government bodies.

More Google coverage

The methodology is public, but the scores are not

The pilot leaves one major question unanswered: how Gemini 2.5 Flash Lite performed. DeepMind’s announcement and technical report describe the evaluation architecture and safety categories but do not publish model scores or a task-by-task results breakdown.

The technical report also acknowledges several limitations. Some proprietary inference code could not be fully inspected or allowlisted, individual Confidential Space builds were not independently reproducible, and Google services were used to sign and verify the attestation report, placing Google in the verification path and increasing the trust required in the model provider.

MLCommons also cautioned that technical secrecy alone is not enough; legal protections and careful benchmark stewardship remain important.

What this could change

The bigger significance of the experiment is not how Gemini scored on one safety benchmark. It is whether AI companies can eventually prove that their benchmark results were earned without allowing evaluators or developers to influence the test.

That distinction could become increasingly important as benchmark scores shape decisions by regulators, researchers and businesses. A secure evaluation process could make independent testing easier without forcing companies to surrender model weights or evaluators to expose valuable test sets.

For IT leaders assessing vendor claims, the method could eventually provide stronger evidence that AI models were tested against independent, previously unseen benchmarks. Until the process becomes reproducible and detailed results are released, buyers should still ask who supplied the benchmark, who evaluated the outputs, what findings were disclosed and which parts of the system required trust in the model provider.

But for double-blind testing to become a meaningful industry standard, the process will need to be independently reproducible, transparent about methodology and capable of scaling across models and benchmarks. Otherwise, the industry could end up with more secure tests without necessarily having more trustworthy results.

Read more: Google’s restricted rollout of Gemini 3.5 Flash Cyber shows why enterprises should examine independent performance evidence and testing controls before adopting specialized AI models.