Google and WhatsApp have confirmed they are aware of a bug that makes it appear as if WhatsApp is accessing phones’ microphones unnecessarily on some Android devices. The issue first cropped up a month ago, but gained new attention after a Twitter engineer tweeted about it in a post that was boosted by Elon Musk.

An image shared by Twitter engineer Foad Dabiri appeared to show that the microphone had been repeatedly running in the background while he wasn’t using the app. He tweeted a screenshot from Android’s Privacy Dashboard, which tracks how often apps access a device’s microphone and camera.

Musk retweeted Dabiri’s post, saying “WhatsApp cannot be trusted.” Incidentally, Musk is known to be a fan of Signal, and has said encrypted direct messages on Twitter could roll out as soon as this month. The company didn’t respond to a request for comment.

In a statement shared on Twitter, WhatsApp suggested it was an Android-related issue, and not a result of inappropriate microphone access by the messaging app “We believe this is a bug on Android that mis-attributes information in their Privacy Dashboard and have asked Google to investigate and remediate,” the company said.

Dabiri is not the first to notice the issue. WhatsApp blogwabetainfohighlighted the bug a month ago, describing it at the time as “a false positive” affecting owners of some Pixel and Samsung devices. They added that restarting the phone may be a possible fix. Meanwhile, Google has said little about what could be causing the discrepancy, but confirmed it’s looking into the matter. "We are aware of the issue and are working closely with WhatsApp to investigate,” a Google spokesperson said in a statement.

It has been more than five years since Beats last refreshed its top-end Studio headphones, but a new model could be on the way. According to 9to5Mac, Apple is “about” to launch a set of Beats Studio Pro headphones. The new model reportedly features a custom Beats chip that promises improved active noise cancellation and transparency mode performance. For the first time, the Studio line may also feature personalized spatial audio. Additionally, 9to5Mac speculates the new model will come with a USB-C port for fast charging.

Visually, the headphones look similar to the current Studio3 model, though it appears Apple has done away with the “Studio” branding found on the side of those headphones. Based on codenames found by 9to5Mac in the internal files for iOS 16.5’s release candidate, Apple collaborated with fashion designer Samuel Ross, best known for starting the clothing label A-Cold-Wall, on the design of the Beats Studio Pro. Images the outlet found in those same files suggest Apple will offer the headphones in four colorways: blue, black, brown and white.

It’s unclear if Apple intends for the Beats Studio Pro to replace the $349 Studio3 headphones, or if the company plans to market them as a more premium offering. According to 9to5, Apple is also working on a set of Studio Buds+. They will reportedly support audio sharing, automatic device switching and Hey Siri integration. The outlet suggests both products will arrive in stores soon.

Japanese developer Game Freak is best known for a little franchise called Pokémon, but throughout the years it has dabbled in other genres, like the strategy title Little Town Hero and the rhythm-based platformer HarmoKnight. Now, the company is betting big on a brand-new action adventure IP codenamed Project Bloom.

Game Freak is teaming up with a Take-Two Interactive publishing label called Private Division. You may not recognize the company by name, but it’s been behind a slew of well-regarded titles throughout the years, like Outer Worlds, Kerbal Space Program, Rollerdrome and OlliOlli World, among others. As for Game Freak, last year’s Pokémon Legends: Arceus proved it could handle open worlds and action-heavy gameplay. The company says Project Bloom is a “bold and tonally different” IP from its prior work.

Not much is known about the game, other than some concept art and a short video announcing the partnership between the two developers. Also, we won’t be playing this anytime soon, as the “sweeping new action-adventure game” isn’t slated for release until Take-Two's 2026 financial year, which runs from April 1, 2025, to March 31, 2026.

It might be surprising to hear that Take-Two Interactive is publishing the game, and not Nintendo, which has been Game Freak's partner on Pokémon for over 25 years. While Pokémon is co-owned by Nintendo, Game Freak is an independent developer. To that end, no gaming console has been mentioned as a home for the forthcoming title. As the game won't be released for three years, the home console landscape is likely to look different than it does now. At the very least, the long-rumored followup to the Switch should be out by then.

Years before everyone was being impressed with the human-like text output of ChatGPT and other generative AI systems, IBM's Watson was blowing our minds on Jeopardy. IBM's cognitive computing project famously dominated its human opponents, but the company had much larger long-term goals, such as using Watson's ability to simulate a human thought process to help doctors diagnose patients and recommend treatments. That didn't work out. Now, IBM is pivoting its supercomputer platform into Watsonx, an AI development studio packed with foundation and open-source models companies can use to train their own AI platforms.

If that sounds familiar, it may be because NVIDIA recently announced a similar service with its AI Foundations program. Both platforms are designed to give enterprises a way to build, train, scale and deploy an AI platform. IBM says Watsonx will provide AI builders with a robust series of training models with an auditable data lineage — ranging from datasets focused on automatically generating code for developers or for handling industry-specific databases to climate datasets designed to help organizations plan for natural disasters.

IBM has already built an example of what the platform can do with that latter dataset in collaboration with NASA, using the geospatial foundation model to convert satellite images into maps that track changes from natural disasters and climate change.

Reimagining Watson as an AI development studio might lack the pizazz of a headline-grabbing supercomputer that can beat humans on a TV quiz show — but the original vision of Watson was out of reach for the average person. Depending on how companies use IBM's new AI training program, you may find yourself interacting with a part of Watson yourself sometime in the near future.

Watsonx is expected to be available in stages, starting with the Watsonx.ai studio in July, and expanding with new features debuting later this year.

A Pokémon-inspired art collection is visiting Los Angeles this summer. The Pokémon × Kogei craftwork exhibition is a collaboration between The Pokémon Company and LA’s Japan House, a space dedicated to spotlighting Japanese culture. The collection includes more than 70 pieces of art crafted by 20 artists, each piece filled to the brim with pocket monsters.

The exhibition focuses on crafts over drawings and paintings, as the art spans multiple mediums like lacquer, ceramics, textiles, metalwork and more. The artists involved with the collection include metalworks legend Morihito Katsura and sculptor Taiichiro Yoshida, among others. The collection is arranged into three sections.

The “Stories” section emphasizes franchise mascot Pikachu. A much-anticipated piece here is called “Pikachu Forest,” which is made from more than 900 strands of lace suspended from the ceiling to create, well, a forest populated by electrified yellow rodents. The “Appearance” section includes many pieces featuring Eevee and its various evolutions, like Jolteon, Vaporeon and Flareon. The final section, called “Life”, seems to be a hodge-podge affair with many different Pokémon.

The exhibition starts on July 25th and runs for five full months, all the way until January. Admission is free and the gallery is open seven days a week. Japan House also says they'll be hosting events related to the exhibition but has not announced any details. In any case, this seems like a great place to meet friends for some Pokémon Go action.

Meta is open-sourcing an AI tool called ImageBind that predicts connections between data similar to how humans perceive or imagine an environment. While image generators like Midjourney, Stable Diffusion and DALL-E 2 pair words with images, allowing you to generate visual scenes based only on a text description, ImageBind casts a broader net. It can link text, images / videos, audio, 3D measurements (depth), temperature data (thermal), and motion data (from inertial measurement units) — and it does this without having to first train on every possibility. It’s an early stage of a framework that could eventually generate complex environments from an input as simple as a text prompt, image or audio recording (or some combination of the three).

You could view ImageBind as moving machine learning closer to human learning. For example, if you’re standing in a stimulating environment like a busy city street, your brain (largely unconsciously) absorbs the sights, sounds and other sensory experiences to infer information about passing cars and pedestrians, tall buildings, weather and much more. Humans and other animals evolved to process this data for our genetic advantage: survival and passing on our DNA. (The more aware you are of your surroundings, the more you can avoid danger and adapt to your environment for better survival and prosperity.) As computers get closer to mimicking animals’ multi-sensory connections, they can use those links to generate fully realized scenes based only on limited chunks of data.

So, while you can use Midjourney to prompt “a basset hound wearing a Gandalf outfit while balancing on a beach ball” and get a relatively realistic photo of this bizarre scene, a multimodal AI tool like ImageBind may eventually create a video of the dog with corresponding sounds, including a detailed suburban living room, the room’s temperature and the precise locations of the dog and anyone else in the scene. “This creates distinctive opportunities to create animations out of static images by combining them with audio prompts,” Meta researchers said today in a developer-focused blog post. “For example, a creator could couple an image with an alarm clock and a rooster crowing, and use a crowing audio prompt to segment the rooster or the sound of an alarm to segment the clock and animate both into a video sequence.”

Meta

As for what else one could do with this new toy, it points clearly to one of Meta’s core ambitions: VR, mixed reality and the metaverse. For example, imagine a future headset that can construct fully realized 3D scenes (with sound, movement, etc.) on the fly. Or, virtual game developers could perhaps eventually use it to take much of the legwork out of their design process. Similarly, content creators could make immersive videos with realistic soundscapes and movement based on only text, image or audio input. It’s also easy to imagine a tool like ImageBind opening new doors in the accessibility space, generating real-time multimedia descriptions to help people with vision or hearing disabilities better perceive their immediate environments.

“In typical AI systems, there is a specific embedding (that is, vectors of numbers that can represent data and their relationships in machine learning) for each respective modality,” said Meta. “ImageBind shows that it’s possible to create a joint embedding space across multiple modalities without needing to train on data with every different combination of modalities. This is important because it’s not feasible for researchers to create datasets with samples that contain, for example, audio data and thermal data from a busy city street, or depth data and a text description of a seaside cliff.”

Meta views the tech as eventually expanding beyond its current six “senses,” so to speak. “While we explored six modalities in our current research, we believe that introducing new modalities that link as many senses as possible — like touch, speech, smell, and brain fMRI signals — will enable richer human-centric AI models.” Developers interested in exploring this new sandbox can start by diving into Meta’s open-source code.

With an instantly recognizable hook and effervescent melody, Vanessa Carlton’s debut single A Thousand Miles hit the 2002 Billboard Charts like a neutron bomb, earning nominations for the Grammy Award for Song of the Year and the Billboard Music Award for Top 40 Track of the Year. Featured prominently in 2004’s White Chicks, Terry Crews credits the undeniable smash with helping launch his acting career. 

The accompanying music video saw Carlton and her piano rolling through Newbury Park, California, and portions of downtown Los Angeles. Twenty-one years later, a team of hobbyist roboticists have brought Carlton’s music back to the public ear — this time, to the streets of San Francisco with an animatronic performer and remotely deployable disco ball.

The robot, which currently doesn’t have much of a moniker from the team beyond “The Robot,” is the brainchild of San Francisco-based aerospace engineer Ben Howard, electrical engineer Noah Klugman, lawyer Lane Powell (with additional assistance from local puppeteer, Adam Kreutinger). “This is just a thing that we've done together, the three of us, to try to create some joy,” Klugman told Engadget during a recent video call.

The trio first collaborated during the pandemic. “Kids couldn't really trick or treat properly,” Howard explained. “So we put together a kind of spooky Halloween candy dispensing robot that could drive around the streets and any kids who were brave enough could walk up, have a conversation with it and get some candy.” That project inspired them to look into developing a robot with year-round appeal. A “piano playing Muppet seemed like a good thing to do,” he continued, and from that the Thousand-Mile Machine was born.

The team started with an outdated food delivery drone model, obtained from “a friend of a friend,” as the mobile platform on which to build out the rest of the construct. “When companies get rid of these things, if they're cool pieces of hardware, there are plenty of engineers around the city who like to modify them and turn them into fun projects,” Howard explained. “There's a big community of people who are sharing cool hardware around.”

“I came to acquire [the wheeled base] and we wanted to do this music playing robot.” he added. “Then, when you think about piano player that roams around the city, immediately that [Vanessa Carlton] video comes to mind. It's so iconic.”

The nearly 400-pound robot measures roughly five feet long on a side and about four feet tall, narrow enough to fit on a sidewalk and into the TEU container workshop in which it was built at San Francisco’s Box Shop. The wheeled base is controlled remotely and manually, while the puppet’s performance — from the hand and head movements to the big disco ball reveal — are all part of a prerecorded act, akin to Chuck E Cheese’s animatronic Pizza Players band. A single button press is all that’s needed to start the performance.

Vanessa Carlton herself reportedly met the robot during a recent event in Petaluma, “it seemed like she enjoyed it,” Klugman noted. “Everyone we've met in San Francisco has seemed to really love it. I think the response has been overwhelmingly positive.”

“That was very much [the case with] everyone we encountered when we were out filming,” Lane added. “Just really happy to watch it and excited to talk to us about it and just 100 percent positive from all ages and all walks of life all over the city. It was a really cool experience.”

Data breaches and security failures happen everyday. There’s little we can do about that if we want to participate in modern society, except maybe switch out the companies we interact with for their competitors if we presume one to be more secure. There’s one service that we don’t have a choice on whether to interact with, no matter how high profile its security incidents become: the federal government.

A breach of the Office of Personnel Management announced in 2015 it had leaked background investigation records, impacting 21.5 million individuals, according to the agency. The highly publicized Solarwinds hack discovered in 2020 exposed government and business records to Russian insiders. Earlier this year, the US Marshals Service division of the Department of Justice became a target, when hackers stole personal information about investigation targets, personnel and more.

The attacks were targeted, usually seeking out some type of sensitive state information. But we all have sensitive information stored throughout federal agencies like our social security numbers or home addresses. Probably even more information is at stake if you utilize federal services like Medicare, student loans or SNAP benefits. We have no choice but to give the federal government access to our personal information in exchange for certain services, unless you’re reading this while living off grid.

“If we want to live in the information age, and we're using some of these systems, we are inherently giving up control,” Kevin Cleary, clinical assistant professor of management science and systems at University at Buffalo, told Engadget. “You have to trust that agency has put forward all the best controls and practices.”

In response, the federal government has developed agencies like the Cybersecurity and Infrastructure Security Agency to lead better security initiatives across departments. In part, this is intended to help you feel a little bit better about storing your data within federal servers by setting higher standards for how it safeguards your data. According to Michael Duffy, associate director of the cybersecurity division at CISA, since the agency’s establishment in 2018, it’s spearheaded the most progress he’s seen in his federal cybersecurity career.

So, things are improving, and you can probably trust the federal government to keep your data safe in the same way you trust the companies you interact with everyday. What makes the government so different, though, is that it’s a high profile target. Adversarial countries want in on state secrets while, at the same time, it’s hard to prioritize spending on security measures. Getting tax-payer funds to fill a pothole on your local highway is hard enough when the damage is tangible and obvious, while security is hard to quantify the benefits of until an attack occurs. In other words, the value of security investments aren’t proven until it’s already too late.

This has gotten better. Security investments in the federal government largely trend upwards. Still, it’s not enough. “Sometimes their budgets don't allow them to take every step or to everything that they would like to do, because you just simply don't have the money,” Marisol Cruz Cain, director of information technology and cybersecurity at GAO, said.

But the reason why the federal government may appear less secure is because of its obligation for transparency. There’s a responsibility to share lessons learned after an incident, and make sure citizens know what happened. That’s actually a big part of CISA’s job. “We are really looking at ways that we are making it more acceptable to raise the hand and say this is the way that we were attacked or an incident occurred,” Duffy said.

The government also interacts with a ton of outside businesses. So, say a government contractor experiences a breach or security incident, that means that data held in federal tech could be exposed. This opens up a slew of new attack vectors, and possibilities for malpractice.

You can actually see how secure certain agencies are thanks to the Government Accountability Office (GAO) and legislation like the Federal Information Technology Acquisition Reform Act. The latter documents tech modernization efforts across major agencies, including cyber readiness. GAO, for its part, audits cybersecurity efforts and develops privacy impact assessments that are publicly available descriptions about what information the agency collects, how they use it and more.

But with all these audits come a relatively bleak conclusion. Agencies aren’t evaluating their policies and procedures to make sure that high profile incidents don’t happen on a regular basis, Cruz Cain said. Your information will be on those servers whether you like it or not.

When it comes to streaming devices, there’s almost too much choice out there. However, if you’re looking for an affordable but capable option, it’s hard to go wrong with Amazon’s Fire TV offerings, especially when those devices are on sale like they are right now.

The highlight of the promotion is the Fire TV Stick 4K Max. Thanks to a $20 discount, you can get the device for $35. The 4K Max is Amazon’s most powerful streaming stick, thanks to a faster processor the company claims is 40 percent more powerful than the one found in the Fire TV Stick 4K. The additional processing power translates to faster app start times and more fluid menu navigation. The Fire TV Sitck 4K Max also comes with WiFi 6 connectivity, as well as support for Dolby Vision, HDR, HDR10+ and Doly Atmos audio.

If you want to add Amazon’s Fire TV interface to an older Full HD TV, consider the Fire TV Stick Lite. It’s down to $20 following a $10 discount. The Fire TV Stick Lite is on Engadget’s list of the best streaming devices you can buy. It might not offer 4K output, but it does come, like its more expensive siblings, with an Alexa Voice Remote out of the box. It also supports HDR output, so you won't miss out on HDR content just because you're buying a more affordable streaming stick.

On the other end of the scale is the latest Fire TV Cube. It’s currently priced at $125, down from $140. Thanks to the inclusion of a 2.0GHz octa-core processor, the Fire TV Cube is the most powerful streaming device Amazon offers. It’s a great option if you own a WiFi 6E-capable router and want to control some of your other entertainment devices with Alexa.

Follow @EngadgetDeals on Twitter and subscribe to the Engadget Deals newsletter for the latest tech deals and buying advice.

It is not hard — at all — to trick today’s chatbots into discussing taboo topics, regurgitating bigoted content and spreading misinformation. That’s why AI pioneer Anthropic has imbued its generative AI, Claude, with a mix of 10 secret principles of fairness, which it unveiled in March. In a blog post Tuesday, the company further explained how its Constitutional AI system is designed and how it is intended to operate.

Normally, when an generative AI model is being trained, there’s a human in the loop to provide quality control and feedback on the outputs — like when ChatGPT or Bard asks you rate your conversations with their systems. “For us, this involved having human contractors compare two responses,” the Anthropic team wrote. “from a model and select the one they felt was better according to some principle (for example, choosing the one that was more helpful, or more harmless).”

Problem with this method is that a human also has to be in the loop for the really horrific and disturbing outputs. Nobody needs to see that, even fewer need to be paid $1.50 an hour by Meta to see that. The human advisor method also sucks at scaling, there simply aren’t enough time and resources to do it with people. Which is why Anthropic is doing it with another AI.

Just as Pinocchio had Jiminy Cricket, Luke had Yoda and Jim had Shart, Claude has its Constitution. “At a high level, the constitution guides the model to take on the normative behavior described [therein],” the Anthropic team explained, whether that’s “helping to avoid toxic or discriminatory outputs, avoiding helping a human engage in illegal or unethical activities, and broadly creating an AI system that is ‘helpful, honest, and harmless.’”

According to Anthropic, this training method can produce Pareto improvements in the AI’s subsequent performance compared to one trained only on human feedback. Essentially, the human in the loop has been replaced by an AI and now everything is reportedly better than ever. “In our tests, our CAI-model responded more appropriately to adversarial inputs while still producing helpful answers and not being evasive,” Anthropic wrote. “The model received no human data on harmlessness, meaning all results on harmlessness came purely from AI supervision.”

The company revealed on Tuesday that its previously undisclosed principles are synthesized from “a range of sources including the UN Declaration of Human Rights, trust and safety best practices, principles proposed by other AI research labs, an effort to capture non-western perspectives, and principles that we discovered work well via our research.”

The company, pointedly getting ahead of the invariable conservative backlash, has emphasized that “our current constitution is neither finalized nor is it likely the best it can be.”

“There have been critiques from many people that AI models are being trained to reflect a specific viewpoint or political ideology, usually one the critic disagrees with,” the team wrote. “From our perspective, our long-term goal isn’t trying to get our systems to represent a specific ideology, but rather to be able to follow a given set of principles.”