“Alexa+ is much more powerful than the original Alexa, which is why we’re working to make it available to as many customers as possible,” an Amazon spokesperson said. “We have notified customers of the upcoming update, including the option to revert to the original version.”

Amazon says it is only upgrading Prime members with supported devices to Alexa+ Early Access as part of this fresh push. Alexa+ is currently free, and Prime members will continue to get it for free after the first full version launches, but it will cost $20 per month for non-Prime members once it officially launches. The list of supported devices is also extensive and includes the first generations of Echo, Echo Dot, Echo Show, Echo Spot and even Echo Plus. There’s a long list of compatible Fire TV and Fire tablets, so depending on how many Amazon-powered devices you have and whether you’re a Prime member, you may see Alexa+ in several places.

How to go back to the original Alexa

If you don’t want to upgrade or don’t like the experience, you can easily revert to the previous version. Just tell your Alexa device, “Alexa, exit Alexa Plus,” and your device will revert to the original Alexa voice and experience. This works on supported Echo devices. You’ll no longer see a chatbot-like conversation style, and your assistant’s responses will become less conversational.

Plus, you won’t really lose your core capabilities. Personally, I haven’t found that I need to change how I ask questions between the two versions of the voice assistant. The on-screen experience and voices have returned to something very similar (if not slightly updated, although I’m using the latest Echo Show 11 to test it) to how I remember Alexa before I switched to Alexa+. Amazon will also warn you that some features of the Alexa app will need to change if you revert to an older version, but it won’t specify what those are or make it basic to find those settings.

If you’re not a Prime member and haven’t received an automatic push to Alexa+, you should still be able to easily switch it in the app or even on your device by asking to switch to Alexa+. There is no longer a registration page like there used to be as there is no waiting list. To find details on how to access Alexa+ and manage your controls, see the Memberships and Subscriptions section of your Amazon account.

Alexa vs. Alexa+

Even if you ask to go back to the ancient Alexa, right Really ancient Alexa? Although it is the same voice, Feminine 2 (or “Relaxed”), our band – together with other users on Reddit— I noticed some changes in the voice, which made it sound different than before. You can also utilize the Feminine 2 with Alexa+, so the voice itself has likely been updated to keep up with the more conversational style that Alexa+ promises.

Some people also hate the fresh default voice, Feminine 1, which the Alexa app describes as “upbeat,” but others call it a “teenage girl” voice. (Or as my editor, a parent of a teenager, calls it, “the sassy voice of a teenager.”) The fresh Alexa sounds much younger and its rhythm sounds almost sarcastic. In total, Alexa+ offers four female and four male voices, each of which is described in the app with one adjective, such as “Inviting” or “Warm”.

Employees in Google DeepMind has asked company management for plans and policies to ensure “physical security” from Immigration and Customs Enforcement (ICE) while on company property, according to screenshots of internal messages obtained by WIRED.

On Monday morning, two days after federal agents shot and killed Minneapolis nurse Alex Pretti, a Google DeepMind employee sent the following message on an internal message board to the company’s approximately 3,000-person artificial intelligence department:

“U.S.-focused question: What is GDM doing to keep us physically safe from ICE? The events of the past week have shown that immigration status, citizenship, and even the law are no deterrent to detention, violence, and even death at the hands of federal agents.”

He continues: “What plans and policies are in place to keep us safe in the office? On the way to and from work? As we have seen, government agency tactics can change and escalate quite quickly. With offices in multiple metropolitan areas across the United States, are we prepared?”

The message received over 20 “plus emoji” reactions from Google DeepMind employees.

As of Monday evening, no senior Google leaders had responded to the news. In fact, Google’s top management – including CEO Sundar Pichai and DeepMind CEO Demis Hassabis – have remained quiet about Pretti’s murder even within the company, sources said.

The news highlights some of the latest divisions forming between artificial intelligence companies and their employees over the Trump administration’s deployment of federal immigration agents to America. While Silicon Valley CEOs have largely bent the knee to Trump, their employees have begun to raise concerns internally and externally about the federal government’s actions.

Google DeepMind’s chief scientist, Jeff Dean, is one of the industry’s most outspoken critics of ICE. In write to X On Sunday, he responded to the video of Pretti’s shooting by saying, “It’s absolutely disgraceful.”

Employees at defense technology company Palantir are questioning the company’s decision to cooperate with ICE. WIRED previously reported that one Palantir employee wrote on Slack: “I think ICE are the bad guys. I’m not proud that the company I love working for so much is a part of it.”

Employees at artificial intelligence labs that work with Palantir – including OpenAI, Google, Anthropic and Meta – also weighed whether to urge leaders to cut ties with the defense technology company. New York Times reported.

Concerns about ICE agents entering Google offices are not unfounded. In a message obtained by WIRED, another Google DeepMind employee expressed concerns about an alleged attempt by a federal agent to enter the company’s Cambridge, Massachusetts, office in the fall.

Google’s chief security and risk officer responded to this message to explain what happened. They noted that “the officer arrived at the reception without notice” and that the agent “was not allowed entry because he did not have a warrant and left immediately.”

Google declined to comment.

Google is one of many Silicon Valley companies that employs thousands of highly skilled foreign workers, many of whom are in the United States on visas. In featherlight of the Trump administration’s immigration crackdown, these companies had something to offer increased security for many of its employees. At the end of last year, Google and Apple advised workers on visas not to leave the country after the White House tightened vetting of visa applicants.

At the time, Silicon Valley leaders were not shy defense of visa programsthat allowed the United States to bring in top talent from around the world.

However, AI executives seem hesitant to comment on the federal government’s latest actions on immigration. Outside of Google, top executives from Silicon Valley companies – including OpenAI, Meta, xAI, Apple and Amazon – have yet to publicly comment on ICE’s actions. According to OpenAI CEO Sam Altman, he referred to the Minnesota incident in an internal message to the company DealBooktelling employees that “what is happening with ICE is going too far.”

Google is giving Photos users more control over its generative photo-to-video AI feature. Google Photos now supports text suggestions for video generation update announcement on Mondayallowing users to describe the specific movement, style or effect they want to see, bringing still images to life.

Text suggestions join the existing “Subtle Motion” and “I’m Lucky” options that are already available in Google Photos’ image-to-video tool. None of them allowed users to enter their own descriptions to support obtain results. Google says Photos will quickly display suggestions “for instant video inspiration,” and text suggestions can be edited to fine-tune the resulting video generations.

The novel text prompt feature is only available to users who are 18 years of age or older. However, this age restriction only applies to Google’s image-to-video tool in Photos – text suggestions have been available for the same features in Gemini for some time, and are available to all users 13 and older. The ability to add descriptive tooltips brings the Photos image-to-video tool more in line with competing AI editors like xAI’s Grok, which has been particularly abused for stripping down photos of real people and children.

Generative AI features in Google Photos are not yet available in all regions, and exact functionality may vary by location. You can test yourself Google support page for more information on current restrictions.

The update also allows Google Photos users to include audio by default in any video they create using the video-to-video feature, making the videos web-ready without additional editing. In another updateGoogle has introduced a novel Google Photos picker for Gmail that makes it easier to share photos and videos from albums, collections, and shared albums via email. This feature allows you to select multiple photos and videos at once and search for specific photos and album names.

Anthropic’s Claude got a little lively today with a recent extension to MCP, an open-source protocol that allows AI agents to easily access tools and data on the Internet. Users will now be able to interact with applications directly within the Claude chatbot, allowing them to create and format Slack messages to colleagues and create presentations for clients in Canva without having to switch tabs.

Starting today, Anthropic says tools like Asana, Figma, Slack, and Canva will “open as interactive apps right in the chat.” While users could previously connect tools like Slack and Asana to the AI ​​assistant, this meant text recovery. The company says the recent in-app integration means users can actually exploit the tool and “see, explore and refine results visually, rather than just reading about them.”

Many popular tools are already available directly from Claude. Anthropic says users can customize Canva decks in real time, format and view messages in Slack, create interactive charts with Hex or Amplitude, and manage projects with Asana or monday.com. Anthropic says Figma, Clay, and Box integrations are also available, and Salesforce tools like Data 360, Agentforce, and Customer 360 are “coming soon.”

The interactivity—sort of like an app within an app—resembles the “mini” apps built into messaging platforms like Telegram and Discord. The integration signals a shift in the way AI platforms like ChatGPT and Claude will likely work in the future, becoming more operating systems than individual tools, creating an “app for everything” like Tencent’s WeChat in China. ChatGPT has already taken a step in this direction by launching its own app ecosystem last year.

Anthropic credits the Model Context Protocol (MCP) with enabling its application integration with Claude. The extension, called MCP Apps, “allows any MCP server to provide an interactive interface within any AI-enabled product – not just Claude,” which means other interactive app interfaces could soon be made available to other AI tools.

Operational standards such as MCP are critical to building useful product ecosystems because they avoid the need for companies to develop and maintain many different interfaces. MCP launched at Anthropic in 2024, but has been widely adopted by companies such as OpenAI, Google and Microsoft. Tardy last year, Anthropic donated it to the Linux Foundation and, along with other tech giants like OpenAI, Google, Microsoft, AWS, Block, Bloomberg, and Cloudflare, created a recent fund — the Agentic AI Foundation — to “advance open-source agentic AI.”

X is facing an investigation by the European Commission over sexualized deepfakes generated by his Grok AI chatbot. IN his announcementThe commission says it will assess whether previously by New York Times.

“The sexual imitation of women and children is a brutal, unacceptable form of degradation,” says Henna Virkkunen, Executive Vice-President of the European Commission for Technological Sovereignty, Security and Democracy, in a press release. “As part of this investigation, we will determine whether X complied with his legal obligations under the DSA or treated the rights of European citizens – including women and children – as collateral damage to his services.”

X could be fined up to 6 percent of its annual worldwide revenue if it is again found to be in violation of the DSA.

Not only did OpenAI co-founder and longtime CEO Greg Brockman make a regular donation to a major pro-Trump super PAC, but his and his wife Anna’s combined donations in September 2025 were the largest of all at $25 million to “MAGA Inc.” last report. The Brockmans’ donations accounted for nearly a quarter of the six-month fundraising cycle.

This is the latest in a string of examples of tech executives cozying up to President Trump’s administration as the administration pushes to aggressively support the artificial intelligence industry and push back on state-level regulation, something that companies like OpenAI largely oppose. OpenAI did not immediately respond to a request for comment.

Brockman’s multimillion-dollar donation is not the only example of him spending immense sums under his own name on lobbying activities aimed at eliminating potential regulations of the artificial intelligence industry. The pro-AI super PAC “Leading the Future,” of which Brockman is a prominent supporter, purchased ads targeting Up-to-date York State Assemblyman Alex Bores, a co-sponsor of Up-to-date York’s RAISE Act, which was watered down at the last minute after a concerted lobbying effort.

Although news of the Brockmans’ donations first broke earlier this month, online discussion has been revived following the recent death of Alex Pretti in Minneapolis, where federal officers fatally shot two people during anti-immigrant attacks. Technical workers across the industry, including many at OpenAI, have been through this signed the letter calling on their CEOs to cancel all contracts with Immigration and Customs Enforcement (ICE) and publicly condemn the department’s actions. “When Trump threatened to send the National Guard to San Francisco in October, tech industry leaders called the White House,” we read on the petition website. “It worked: Trump backed down. Today we are calling on our CEOs to pick up the phone again.”

But since Trump’s inauguration, more tech leaders have donated to the inaugural fund, traveled to Mar-a-Lago to meet him, or attended White House dinners by his side. Instead, they received an administration willing to roll back consumer protections and technical regulations. Trump’s Artificial Intelligence Action Plan resurrected a failed Republican attempt to bar states from passing AI laws, targeting tech leaders delight. The modern rule states that “AI is too critical to be stifled by bureaucracy at this early stage” and that the government “should not allow AI-related federal funds to be directed to states with burdensome AI regulations that waste those funds,” although it should also not “interfere with states’ right to enact careful regulations that do not unduly restrict innovation.” The moratorium includes SB 53, the landmark artificial intelligence transparency bill signed by California Governor Gavin Newsom in September despite lobbying against it by many tech companies, including OpenAI.

In 2019, Brockman co-authored, among others: blog post about how challenging it is to “change powerful systems… once they are implemented” and that “it is critical to address AGI security and policy risks before they are created.” Six years later, his posts have changed their tone, emphasizing the importance of “approach.”[ing] new technologies with a focus on development.” In the Up-to-date Year’s Eve post on XBrockman wrote that “this year, my wife Anna and I began to engage politically, including through political contributions, which reflected support for policies that support American innovation and constructive dialogue between government and the technology sector.” He added that it was “great to see the willingness of the president and his administration to engage directly with the AI ​​community.”

Microsoft today announces the successor to its first proprietary AI chip, the Maia 200. Built on TSMC’s 3nm process, Microsoft claims its Maia 200 AI accelerator “delivers 3x faster fourth-training performance compared to third-generation Amazon Trainium and eighth-training performance superior to Google’s seventh-generation TPU.”

Each Maia 200 chip has over 100 billion transistors, designed to handle large-scale AI workloads. “Maia 200 can easily support today’s largest models, leaving plenty of room for even larger models in the future,” says Scott Guthrie, executive vice president of Microsoft’s Cloud and AI division.

Microsoft will exploit Maia 200 to host OpenAI’s GPT-5.2 model and others for Microsoft Foundry and Microsoft 365 Copilot. “Maia 200 is also the most efficient inference system Microsoft has ever deployed, delivering 30 percent better performance per dollar compared to the next-generation hardware currently in our fleet,” says Guthrie.

Microsoft’s performance gap against close Gigantic Tech competitors is different from when Microsoft first released Maia 100 in 2023 and didn’t want to get drawn into head-to-head comparisons with Amazon and Google’s AI cloud capabilities. However, both Google and Amazon are working on next-generation AI chips. Amazon is equal works with Nvidia to integrate the upcoming Trainium4 chip with NVLink 6 and Nvidia’s MGX rack architecture.

Microsoft’s Superintelligence team will be the first to exploit Maia 200 chips, and the company is also inviting researchers, developers, AI labs, and open source modelers to an early preview of the Maia 200 development kit. Microsoft is starting to deploy these modern chips today in the US Azure data center region, with additional regions to follow.

Last month, Pamela Griffin and two other Taylor, Texas residents took to the podium at a city council meeting to express opposition to the data center project. But they later sat down as council members discussed the proposed technology factory. Griffin did not speak out against the measure. Nobody did it.

A similar contrast is echoed in communities across the United States. Data centers are facing unprecedented public resistance, with environmental costs being a major concern. More of them were needed to meet the growing appetite for AI, and they became obvious flashpoints for communities concerned about what automation might mean for them. But many of the factories being built to supply servers, electrical equipment and other parts to data centers face virtually no opposition.

Factories typically create more jobs and employ fewer natural resources than data centers, so with the exception of a few controversial chip factories in several states, they participate in local hearings to obtain permits and tax breaks. But experts who track supply chains say the minimal scrutiny of manufacturing projects highlights a potential fresh strategy for anti-data center activists and a source of risk for communities that may invest in a short-term boom.

“At some point, people will figure out what is the most important factory that can bring all the data centers to their knees, and they will move on,” says Andy Tsay, a professor at Santa Clara University who studies global trade and reshoring.

While focusing on the supply chain may be a fresh way to snail-paced data center construction, Griffin says organizers are too spread out to take on more. So for now, the door is wide open for manufacturers to expand their U.S. presence and feed into the data center market without overwhelming resistance.

“We have to start at the bottom and get the guys who make these servers, but first we have to convince people what these data centers are,” Griffin says. “We have to choose our battles.”

At last month’s council meeting, she focused on opposition to a proposal to build a second data center in Taylor after one was built near her home, which she asks to stop. That evening, Griffin and her colleagues knew the council would also consider a proposed plant for Taiwanese manufacturer Compal. However, the facility’s potential role in supporting the data center industry was not obvious to them.

Griffin’s case highlights what anti-data center communities face when also considering demanding manufacturing projects: opacity, public perception and the prospect of additional legal battles.

Server farms

City documents describe Compal’s intentions as producing “servers” in addition to everything from sharp home devices to car electronics.

It’s a long list, but Compal spokeswoman Tina Chang tells WIRED that the Taylor factory will be dedicated to the company’s server business. The building is leased by a subsidiary of Compal USA Technology which was created last year to expand Compal’s server products business in the US. Another site in nearby Georgetown, Texas, announced at the same time as the Taylor facility, “will establish a server services center to support enterprise and cloud infrastructure needs.” according to the company.

Taylor, located near Austin, courted Compal for more than a year, which considered alternatives around the world before selecting the city. The 366,000-square-foot pre-built facility convinced the company, which said it was signing the contract lease valued at almost $66 million with plans to invest a total of $200 million. “They fell in love with the openness,” Ben White, president of the Taylor Economic Development Corporation, told the city council at its December meeting. “It gave them the flexibility to do what needed to be done.”

If this all sounds familiar, it’s because Gemini already offers the option to connect to the Workspace app. However, this required more work on the user’s part – I always found that I had to explicitly ask them to check something in my email or calendar if I wanted them to operate them as sources. Now, if a prompt seems like it deserves a trip to your inbox to look for a concert ticket email, it will do it on its own. This is something huge. If you have to be specific with every command and have artificial intelligence take care of it, it won’t get any more useful than the timer-setting robot assistants we’ve been using for the last decade.

Gemini offers some suggested prompts you can try once you turn on Personal Intelligence, such as recommending books you might like based on your interests. The titles he suggested to me were annoyingly correct. Another one of these conversation starters led to a long chat about strategies for dealing with the lawn in my yard, which I hate and the crows are tearing it apart anyway. Gemini suggested some native plants to consider, added reminders to my calendar based on the plan I chose, and prepared a shopping list at Keep that I could take to the hardware store. Up until a few months ago, Gemini routinely failed when I asked it to complete tasks like “Add this to my calendar,” so this is a significant step forward.

The thing is, Gemini handles skiing in a different way. I asked him to brainstorm ideas for fresh bike routes and asked him to include a stop at a coffee shop. He obliged and his high-level recommendations were good, although he struggled with the minutiae. Trying to map out specific routes was painful; would give me a link to the route she supposedly created on Google Maps, but clicking on the map showed me a different set of directions. I’m also not sure about the plan to send me through the forest on unpaved trails, ending with a left turn crossing several lanes of traffic on a busy road, so I’ll probably stick to the routes I know.

This is a problem. Gemini can analyze my interests and make some pretty good guesses about what I’ll be interested in; these are the details where artificial intelligence gets lost. I asked him to look up neighborhoods I might be less familiar with and recommend them for an afternoon tour to take photos and (of course) grab a coffee. He used my personal information to correctly determine that I had previously lived in Ballard and that I should not provide this as a recommendation. The overall list he has put together is solid; the specific locations he recommended were not always appropriate.

She claimed that the South Park restaurant was in Georgetown, that I could find Caffe Umbria in the Senior Rainier Brewery building (it’s not there), and strongly endorsed the T-shirt shop, which is quite clearly closed based on the entry on Google Maps. I had to fact-check and repeat enough that it all started to seem like more work than it was worth.

This may be Gemini’s biggest immediate challenge. A year ago I needed a lot of babysitting to get to the personal information I needed, AND Something broke regularly. Now it can reliably perform personal tasks – but getting the details wrong is a pretty massive mistake. You only need to show up to an empty store once to decide you’re done with Gemini. That doesn’t even touch on the privacy aspect of all this. In one of our conversations, Gemini mentioned my husband and child by name. It’s one thing to know that this information is dead uncomplicated to find with access to my email and calendar; it’s another thing to hear their names out deafening.

Disputes aside, I think the inclusion of Personal Intelligence has increased the scope of what I will operate Gemini for – but only slightly, and I didn’t operate it much on a daily basis at first. I have a gardening schedule and a list to take to a local nursery where I’ll ask a real human being if I’m on the right track. Perhaps pre-planning with Gemini will support me feel confident enough to get started, even if I end up course-correcting in the future. It’s not a bad tool. But you can bet I’ll be watching my steps on the path he tells me.