This week Google "unveiled a set of tools for retailers that helps them roll out AI agents," reports the Wall Street Journal, The new retail AI agents, which help shoppers find their desired items, provide customer support and let people order food at restaurants, are part of what Alphabet-owned Google calls Gemini Enterprise for Customer Experience. Major retailers, including home improvement giant Lowe's, the grocer Kroger and pizza chain Papa Johns say they are already using Google's tools to help prepare for the incoming wave of AI-assisted shopping and ordering...

Kicking off the race among tech giants to get ahead of this shift, OpenAI released its Instant Checkout feature last fall, which lets users buy stuff directly through its chatbot ChatGPT. In January, Microsoft announced a similar checkout feature for its Copilot chatbot. Soon after OpenAI's release last year, Walmart said it would partner with OpenAI to let shoppers buy its products within ChatGPT.
But that's just the beginning, reports the New York Times, with hundreds of start-ups also vying for the attention of retailers: There are A.I. start-ups that offer in-store cameras that can detect a customer's age or gender, robots that manage shelves on their own and headsets that give store workers access to product information in real time... The scramble to exploit artificial intelligence is happening across the retail spectrum, from the highest echelons of luxury goods to the most pragmatic of convenience stores.

7-Eleven said it was using conversational A.I. to hire staff at its convenience stores through an agent named Rita (Recruiting Individuals Through Automation). Executives said that they no longer had to worry about whether applicants would show up to interviews and that the system had reduced hiring time, which had taken two weeks, to less than three days.
The article notes that at the National Retail Federation conference, other companies showing their AI advancements included Applebee's, IHOP, the Vitamin Shoppe, Urban Outfitters, Rag & Bone, Kendra Scott, Michael Kors and Philip Morris.

In December 2024 the Google Threat Intelligence Group published research on the code of the commercial spyware "Predator". But there's now been new research by Jamf (the company behind a mobile device management solution) showing Predator is more dangerous and sophisticated than we realized, according to SecurityWeek.

Long-time Slashdot reader wiredmikey writes: The new research reveals an error taxonomy that reports exactly why deployments fail, turning black boxes into diagnostic events for threat actors. Almost exclusively marketed to and used by national governments and intelligence agencies, the spyware also detects cybersecurity tools, suppresses forensics evidence, and has built-in geographic restrictions.

Ars Technica reports: Security firm Mandiant [part of Google Cloud] has released a database that allows any administrative password protected by Microsoft's NTLM.v1 hash algorithm to be hacked in an attempt to nudge users who continue using the deprecated function despite known weaknesses.... a precomputed table of hash values linked to their corresponding plaintext. These generic tables, which work against multiple hashing schemes, allow hackers to take over accounts by quickly mapping a stolen hash to its password counterpart... Mandiant said it had released an NTLMv1 rainbow table that will allow defenders and researchers (and, of course, malicious hackers, too) to recover passwords in under 12 hours using consumer hardware costing less than $600 USD. The table is hosted in Google Cloud. The database works against Net-NTLMv1 passwords, which are used in network authentication for accessing resources such as SMB network sharing.

Despite its long- and well-known susceptibility to easy cracking, NTLMv1 remains in use in some of the world's more sensitive networks. One reason for the lack of action is that utilities and organizations in industries, including health care and industrial control, often rely on legacy apps that are incompatible with more recently released hashing algorithms. Another reason is that organizations relying on mission-critical systems can't afford the downtime required to migrate. Of course, inertia and penny-pinching are also causes.

"By releasing these tables, Mandiant aims to lower the barrier for security professionals to demonstrate the insecurity of Net-NTLMv1," Mandiant said. "While tools to exploit this protocol have existed for years, they often required uploading sensitive data to third-party services or expensive hardware to brute-force keys."
"Organizations that rely on Windows networking aren't the only laggards," the article points out. "Microsoft only announced plans to deprecate NTLMv1 last August."

Thanks to Slashdot reader joshuark for sharing the news.

OpenAI said Friday that it will begin testing ads on ChatGPT in the coming weeks, as the $500 billion startup seeks new revenue streams to fund its continued expansion and compete against rivals Google and Anthropic. The company had previously resisted embedding ads into its chatbot, citing concerns that doing so could undermine the trustworthiness and objectivity of responses.

The ads will appear at the bottom of ChatGPT answers on the free tier and the $8-per-month ChatGPT Go subscription in the U.S., showing only when relevant to the user's query. Pro, Business, and Enterprise subscriptions will remain ad-free. OpenAI expects to generate "low billions" of dollars from advertising in 2026, FT reported, and more in subsequent years. The revenue is intended to help fund roughly $1.4 trillion in computing commitments over the next decade. The company said it will not show ads to users under 18 or near sensitive topics like health, mental health, or politics.

theodp writes: Code.org, the nonprofit backed by AI giants Microsoft, Google and Amazon and whose Hour of AI and free AI curriculum aim to make world's K-12 schoolchildren AI literate, points job seekers to its AI Use Policy in Hiring, which promises dire consequences for those who use AI during interviews or take home assignments without its OK.

Explaining "What's Not Okay," Code.org writes: "While we support thoughtful use of AI, certain uses undermine fairness and honesty in the hiring process. We ask that candidates do not [...] use AI during interviews and take-home assignments without explicit consent from the interview team. Such use goes against our values of integrity and transparency and will result in disqualification from the hiring process."

Interestingly, Code.org CEO Partovi last year faced some blowback from educators over his LinkedIn post that painted schools that police AI use by students as dinosaurs. Partovi wrote, "Schools of the past define AI use as 'cheating.' Schools of the future define AI skills as the new literacy. Every desk-job employer is looking to hire workers who are adept at AI. Employers want the students who are best at this new form of 'cheating.'"

Iran has imposed one of the longest nationwide internet shutdowns in its history, cutting more than 92 million people off from connectivity for over a week as mass anti-government protests continue. TechCrunch reports: As of this writing, Iranians have not been able to access the internet for more than 170 hours. The previous longest shutdowns in the country lasted around 163 hours in 2019, and 160 hours in 2025, according to Isik Mater, the director of research at NetBlocks, a web monitoring company that tracks internet disruptions.

Mater said that the current shutdown in Iran is the third longest on record, after the internet shutdown in Sudan in mid-2021 that lasted around 35 days, followed by the outage in Mauritania in July 2024, which lasted 22 days. "Iran's shutdowns remain among the most comprehensive and tightly enforced nationwide blackouts we've observed, particularly in terms of population affected," Mater told TechCrunch.

The exact ranking depends on how each organization measures a shutdown. Zach Rosson, a researcher who studies internet disruptions at the digital rights nonprofit Access Now, told TechCrunch that according to its data, the ongoing shutdown in Iran is on a path to crack the top 10 longest shutdowns in history. Further reading: Iran Shuts Down Musk's Starlink For First Time

The powerful data privacy watchdog in Italy long known for aggressively policing U.S. and Chinese AI giants is under investigation for possible corruption and embezzlement. Reuters reports: Rome prosecutors are investigating the agency's president, Pasquale Stanzione, and three other board members over alleged excessive spending and possible corruption behind its decisions, Italian news agencies including ANSA as well as the judicial source, who did not wish to be named, said. Stanzione, when asked by reporters to comment on the investigation, said he was "absolutely serene."

The opposition 5-Star Movement said the agency's credibility had been undermined and called for Stanzione to resign. Stanzione declined to answer when asked repeatedly by reporters whether he would step down. The data privacy authority, known in Italy as the Garante, is one of the European Union's most proactive regulators in assessing AI platform compliance with the bloc's data privacy regime. It frequently takes initiatives -- such as requesting information or imposing fines or bans -- on matters affecting high-tech multinationals operating in the country.

Wikipedia turns 25 today, and the online encyclopedia is celebrating that with an announcement that it has signed new licensing deals with a slate of major AI companies -- Amazon, Microsoft, Meta Platforms, Perplexity and Mistral AI. The deals allow these companies to access Wikipedia content "at a volume and speed designed specifically for their needs." The Wikimedia Foundation did not disclose financial terms.

Google had already signed on as one of the first enterprise customers back in 2022. The agreements follow the Wikimedia Foundation's push last year for AI developers to pay for access through its enterprise platform. The foundation said human traffic had fallen 8% while bot visits -- sometimes disguised to evade detection -- were heavily taxing its servers.

Wikipedia founder Jimmy Wales said he welcomes AI training on the site's human-curated content but that companies "should probably chip in and pay for your fair share of the cost that you're putting on us." The site remains the ninth most visited on the internet, hosting more than 65 million articles in 300 languages maintained by some 250,000 volunteer editors.

After widespread backlash over its 2022 decision to remove JPEG-XL support, Google has quietly restored the image format in the latest Chrome/Chromium codebase. Phoronix reports: Back in December they merged jxl-rs as a pure Rust-based JPEG-XL image decoder from the official libjxl organization. At the end of December they did more JPEG-XL plumbing with the enums and build flags for the support. Now as of yesterday they wired up the JXL decoder! The jxl-rs-powered JPEG-XL image decoding is gated by the enable_jxl_decoder build flag but it's enabled by default.

Linus Torvalds has started experimenting with vibe coding, using Google's Antigravity AI to generate parts of a small hobby project called AudioNoise. "In doing so, he has become the highest-profile programmer yet to adopt this rapidly spreading, and often mocked, AI-driven programming," writes ZDNet's Steven Vaughan-Nichols. Fro the report: [I]t's a trivial program called AudioNoise -- a recent side project focused on digital audio effects and signal processing. He started it after building physical guitar pedals, GuitarPedal, to learn about audio circuits. He now gives them as gifts to kernel developers and, recently, to Bill Gates.

While Torvalds hand-coded the C components, he turned to Antigravity for a Python-based audio sample visualizer. He openly acknowledges that he leans on online snippets when working in languages he knows less well. Who doesn't? [...] In the project's README file, Torvalds wrote that "the Python visualizer tool has been basically written by vibe-coding," describing how he "cut out the middle-man -- me -- and just used Google Antigravity to do the audio sample visualiser." The remark underlines that the AI-generated code met his expectations well enough that he did not feel the need to manually re-implement it. Further reading: Linus Torvalds Says Vibe Coding is Fine For Getting Started, 'Horrible Idea' For Maintenance

22 years ago, developer and columnist John Gruber released Markdown, a simple plain-text formatting system designed to spare writers the headache of memorizing arcane HTML tags. As technologist Anil Dash writes in a long piece, Markdown has since embedded itself into nearly every corner of modern computing.

Aaron Swartz, then seventeen years old, served as the beta tester before its quiet March 2004 debut. Google eventually added Markdown support to Docs after more than a decade of user requests; Microsoft put it in Notepad; Slack, WhatsApp, Discord, and Apple Notes all support it now. Dash writes: The part about not doing this stuff solely for money matters, because even the most advanced LLM systems today, what the big AI companies call their "frontier" models, require complex orchestration that's carefully scripted by people who've tuned their prompts for these systems through countless rounds of trial and error. They've iterated and tested and watched for the results as these systems hallucinated or failed or ran amok, chewing up countless resources along the way. And sometimes, they generated genuinely astonishing outputs, things that are truly amazing to consider that modern technology can achieve. The rate of progress and evolution, even factoring in the mind-boggling amounts of investment that are going into these systems, is rivaled only by the initial development of the personal computer or the Internet, or the early space race.

And all of it -- all of it -- is controlled through Markdown files. When you see the brilliant work shown off from somebody who's bragging about what they made ChatGPT generate for them, or someone is understandably proud about the code that they got Claude to create, all of the most advanced work has been prompted in Markdown. Though where the logic of Markdown was originally a very simple version of "use human language to tell the machine what to do", the implications have gotten far more dire when they use a format designed to help expresss "make this **bold**" to tell the computer itself "make this imaginary girlfriend more compliant".

Apple has struck a multi-year partnership with Google to power a more capable version of Siri using Gemini AI models, ending months of speculation about which company would help the iPhone maker catch up in the generative AI race. In a statement, Apple said it had determined after "careful evaluation" that "Google's technology provides the most capable foundation for Apple Foundation Models."

The deal comes after Apple delayed its planned Siri AI upgrade last March, acknowledging that the project was taking "longer than we thought." Bloomberg had reported in August that Apple was in early talks with Google about using a custom Gemini model. Apple also explored potential partnerships with OpenAI, Anthropic and Perplexity, and CEO Tim Cook has said the company plans to integrate with more AI companies over time. The upgraded Siri is expected to perform actions on users' behalf and understand personal context.

"The clock is ticking" on the Hubble Space Telescope, writes the space news site Daily Galaxy, citing estimates from the unofficial "Hubble Reentry Tracker" site (which uses orbital data from the site space-track.org, created by tech integrator SAIC): While Hubble was initially launched into low Earth orbit at an altitude of around 360 miles, it has since descended to approximately 326 miles, and it continues to fall... "The solar flux levels are currently longer in duration and more elevated than previously anticipated, resulting in an earlier reentry forecast for the Hubble Space Telescope if no reboost mission is conducted," Hubble Reentry Trackersays the Hubble Reentry Tracker... ["Hubble has been reboosted three times in its history," the site points out, "all by servicing missions using the Space Shuttle."]

NASA partnered with SpaceX in 2022 to explore the feasibility of raising Hubble to its original altitude of 373 miles. Such an adjustment would have bought Hubble a few more years in orbit. However, the future of this plan remains uncertain, as NASA has not made any official announcements to move forward with it... Solar flux levels, which determine atmospheric drag, have increased in recent years, accelerating the telescope's decline. This change in solar behavior means that the possibility of Hubble reentering Earth's atmosphere in the next five to six years is quite high if no corrective action is taken. ["But it is difficult to estimate this value due to the variability of future solar flux," the site cautions. "In the best case, Hubble may not reenter for 15 more years, around 2040. In the worst case, it could reenter in 4 years..."] Once Hubble reaches an altitude of 248 miles, it is expected that it will have less than a year before reentry...

While Hubble's end may be near, there is a promising new project on the horizon: Lazuli, a privately-funded space telescope funded by former Google CEO Eric Schmidt. Lazuli aims to become the first privately-funded space telescope, and it could be the successor Hubble enthusiasts have been hoping for. Schmidt Sciences, the organization behind the telescope, plans to launch Lazuli by 2028, providing a more modern alternative to Hubble with a larger mirror and enhanced capabilities. The telescope's proposed design includes a 94-inch-wide mirror, which is a significant upgrade from Hubble's 94.5-inch mirror, and will feature updated instruments to capture more detailed data than ever before.

Alphabet-owned Wing "is expanding its drone delivery service to an additional 150 Walmart stores across the U.S.," reports Axios: [T]he future is already here if you live in Dallas — where some Walmart customers order delivery by Wing three times a week. By the end of 2026, some 40 million Americans, or about 12 percent of the U.S. population, will be able to take advantage of the convenience, the companies claim... Once the items are picked and packed in a small cardboard basket, they are loaded onto a drone inside a fenced area in the Walmart parking lot. Drones fly autonomously to the designated address, with human pilots monitoring each flight from a central operations hub....

For now, Wing deliveries are free. "The goal is to expose folks to the wonders of drone delivery," explains Wing's chief business officer, Heather Rivera... Over time, she said Wing expects delivery fees to be comparable to other delivery options, but faster and more convenient. Service began recently in Atlanta and Charlotte, and it's coming soon to Los Angeles, Houston, Cincinnati, St. Louis, Miami and other major U.S. cities to be announced later, according to the article. "By 2027, Walmart and Wing say they'll have a network of more than 270 drone delivery locations nationwide."

Walmart also announced a new deal today with Google's Gemini, allowing customers to purchase Walmart products from within Gemini. (Walmart announced a similar deal for ChatGPT in October.)

Slashdot reader BrianFagioli calls this "a defensive angle that Walmart does not quite say out loud." As AI models answer more questions directly, retailers risk losing customers before they ever hit a website. If Gemini recommends a product from someone else first, Walmart loses the sale before it starts. By planting itself inside the AI, Walmart keeps a seat at the table while the internet shifts under everyone's feet.

Google clearly benefits too. Gemini gets a more functional purpose than just telling you how to boil pasta or summarize recipes. Now it can carry someone from the moment they wonder what they need to the moment the order is placed. That makes the assistant stickier and a bit more practical than generic chat. Walmart's incoming CEO John Furner says the company wants to shape this new pattern instead of being dragged into it later. Sundar Pichai calls Walmart an early partner in what he sees as a broader wave of agent style commerce, where AI starts doing the errands people used to handle themselves.
The article concludes "This partnership serves as a snapshot of where retail seems to be heading..."

A new study "compared how well top AI systems and human workers did at hundreds of real work assignments," reports the Washington Post.

They add that at least one example "illustrates a disconnect three years after the release of ChatGPT that has implications for the whole economy." AI can accomplish many impressive tasks involving computer code, documents or images. That has prompted predictions that human work of many kinds could soon be done by computers alone. Bentley University and Gallup found in a survey [PDF] last year that about three-quarters of Americans expect AI to reduce the number of U.S. jobs over the next decade. But economic data shows the technology largely has not replaced workers.

To understand what work AI can do on its own today, researchers collected hundreds of examples of projects posted on freelancing platforms that humans had been paid to complete. They included tasks such as making 3D product animations, transcribing music, coding web video games and formatting research papers for publication. The research team then gave each task to AI systems such as OpenAI's ChatGPT, Google's Gemini and Anthropic's Claude. The best-performing AI system successfully completed only 2.5 percent of the projects, according to the research team from Scale AI, a start-up that provides data to AI developers, and the Center for AI Safety, a nonprofit that works to understand risks from AI. "Current models are not close to being able to automate real jobs in the economy," said Jason Hausenloy, one of the researchers on the Remote Labor Index study...

The results, which show how AI systems fall short, challenge predictions that the technology is poised to soon replace large portions of the workforce... The AI systems failed on nearly half of the Remote Labor Index projects by producing poor-quality work, and they left more than a third incomplete. Nearly 1 in 5 had basic technical problems such as producing corrupt files, the researchers found.
One test involved creating an interactive dashboard for data from the World Happiness Report, according to the article. "At first glance, the AI results look adequate. But closer examination reveals errors, such as countries inexplicably missing data, overlapping text and legends that use the wrong colors — or no colors at all."

The researchers say AI systems are hobbled by a lack of memory, and are also weak on "visual" understanding.

This week Meta announced several new features for "Meta Ray-Ban Display" smartglasses:

- A new teleprompter feature for the smart glasses (arriving in a phased rollout)

- The ability to send messages on WhatsApp and Messenger by writing with your finger on any surface. (Available for those who sign up for an "early access" program).

- "Pedestrian navigation" for 32 cities. ("The 28 cities we launched Meta Ray-Ban Display with, plus Denver, Las Vegas, Portland, and Salt Lake City," and with more cities coming soon.)


But they also warned Meta Ray-Ban Display "is a first-of-its-kind product with extremely limited inventory," saying they're delaying international expansion of sales due to inventory constraints — and also due to "unprecedented" demand in the U.S. CNBC reports: "Since launching last fall, we've seen an overwhelming amount of interest, and as a result, product waitlists now extend well into 2026," Meta wrote in a blog post. Due to "limited" inventory, the company said it will pause plans to launch in the U.K., France, Italy and Canada early this year and concentrate on U.S. orders as it reassesses international availability...

Meta is one of several technology companies moving into the smart glasses market. Alphabet announced a $150 million partnership with Warby Parker in May and ChatGPT maker OpenAI is reportedly working on AI glasses with Apple.

Yes, a federal judge blocked an attempt by Texas at an app store age-verification law. But this year Silicon Valley giants including Google and Apple "are expected to fight hard against similar legislation," reports Politico, "because of the vast legal liability it imposes on app stores and developers." In Texas, Utah and Louisiana, parent advocates have linked up with conservative "pro-family" groups to pass laws forcing mobile app stores to verify user ages and require parental sign-off. If those rules hold up in court, companies like Google and Apple, which run the two largest app stores, would face massive legal liability... California has taken a different approach, passing its own age-verification law last year that puts liability on device manufacturers instead of app stores. That model has been better received by the tech lobby, and is now competing with the app-based approach in states like Ohio. In Washington D.C., a GOP-led bill modeled off of Texas' law is wending its way through Capitol Hill. And more states are expected to join the fray, including Michigan and South Carolina.

Joel Thayer, president of the conservative Digital Progress Institute and a key architect of the Texas law, said states are only accelerating their push. He explicitly linked the age-verification debate to AI, arguing it's "terrifying" to think companies could build new AI products by scraping data from children's apps. Thayer also pointed to the Trump administration's recent executive order aimed at curbing state regulation of AI, saying it has galvanized lawmakers. "We're gonna see more states pushing this stuff," Thayer said. "What really put fuel in the fire is the AI moratorium for states. I think states have been reinvigorated to fight back on this."
He told Politico that the issue will likely be decided by America's Supreme Court, which in June upheld Texas legislation requiring age verification for online content. Thayer said states need a ruling from America's highest court to "triangulate exactly what the eff is going on with the First Amendment in the tech world.

"They're going to have to resolve the question at some point."

An anonymous reader quotes a report from Ars Technica: Search engine optimization, or SEO, is a big business. While some SEO practices are useful, much of the day-to-day SEO wisdom you see online amounts to superstition. An increasingly popular approach geared toward LLMs called "content chunking" may fall into that category. In the latest installment of Google's Search Off the Record podcast, John Mueller and Danny Sullivan say that breaking content down into bite-sized chunks for LLMs like Gemini is a bad idea.

You've probably seen websites engaging in content chunking and scratched your head, and for good reason -- this content isn't made for you. The idea is that if you split information into smaller paragraphs and sections, it is more likely to be ingested and cited by gen AI bots like Gemini. So you end up with short paragraphs, sometimes with just one or two sentences, and lots of subheads formatted like questions one might ask a chatbot.

According to Google's Danny Sullivan, this is a misconception, and Google doesn't use such signals to improve ranking. "One of the things I keep seeing over and over in some of the advice and guidance and people are trying to figure out what do we do with the LLMs or whatever, is that turn your content into bite-sized chunks, because LLMs like things that are really bite size, right?" said Sullivan. "So... we don't want you to do that."

The conversation, which begins around the podcast's 18-minute mark, goes on to illustrate the folly of jumping on the latest SEO trend. Sullivan notes that he has consulted engineers at Google before making this proclamation. Apparently, the best way to rank on Google continues to be creating content for humans rather than machines. That ensures long-term search exposure, because the behavior of human beings -- what they choose to click on -- is an important signal for Google.

An anonymous reader quotes a report from TorrentFreak: Italy's communications regulator AGCOM imposed a record-breaking 14.2 million-euro fine on Cloudflare after the company failed to implement the required piracy blocking measures. Cloudflare argued that filtering its global 1.1.1.1 DNS resolver would be "impossible" without hurting overall performance. AGCOM disagreed, noting that Cloudflare is not necessarily a neutral intermediary either.

[...] "The measure, in addition to being one of the first financial penalties imposed in the copyright sector, is particularly significant given the role played by Cloudflare" AGCOM notes, adding that Cloudflare is linked to roughly 70% of the pirate sites targeted under its regime. In its detailed analysis, the regulator further highlighted that Cloudflare's cooperation is "essential" for the enforcement of Italian anti-piracy laws, as its services allow pirate sites to evade standard blocking measures.

Cloudflare has strongly contested the accusations throughout AGCOM's proceedings and previously criticized the Piracy Shield system for lacking transparency and due process. While the company did not immediately respond to our request for comment, it will almost certainly appeal the fine. This appeal may also draw the interest of other public DNS resolvers, such as Google and OpenDNS. AGCOM, meanwhile, says that it remains fully committed to enforcing the local piracy law. The regulator notes that since the Piracy Shield started in February 2024, 65,000 domain names and 14,000 IP addresses were blocked.

YouTube is updating search filters so users can explicitly choose between Shorts and long-form videos. The change also replaces view-count sorting with a new "Popularity" filter and removes underperforming options like "Sort by Rating." The Verge reports: Right now, a filter-less search shows a mix of longform and short form videos, which can be annoying if you just want to see videos in one format or the other. But in the new search filters, among other options, you can pick to see "Videos," which in my testing has only showed a list of longform videos, or "Shorts," which just shows Shorts.

YouTube is also removing the "Upload Date - Last Hour" and "Sort by Rating" filters because they "were not working as expected and had contributed to user complaints." The company will still offer other "Upload Date" filters, like "Today," "This week," "This Month," and "This Year," and you can also find popular videos with the new "Popularity" filter, which is replacing the "View count" sort option. (With the new "Popularity" filter, YouTube says that "our systems assess a video's view count and other relevance signals, such as watch time, to determine its popularity for that specific query.")