"Amazon suggested its engineers eschew AI code generation tools from third-party companies in favor of its own ," reports Reuters, "a move to bolster its proprietary Kiro service, which it released in July, according to an internal memo viewed by Reuters." In the memo, posted to Amazon's internal news site, the company said, "While we continue to support existing tools in use today, we do not plan to support additional third party, AI development tools.

"As part of our builder community, you all play a critical role shaping these products and we use your feedback to aggressively improve them," according to the memo.

The guidance would seem to preclude Amazon employees from using other popular software coding tools like OpenAI's Codex, Anthropic's Claude Code, and those from startup Cursor. That is despite Amazon having invested about $8 billion into Anthropic and reaching a seven-year $38 billion deal with OpenAI to sell it cloud-computing services..."To make these experiences truly exceptional, we need your help," according to the memo, which was signed by Peter DeSantis, senior vice president of AWS utility computing, and Dave Treadwell, senior vice president of eCommerce Foundation. "We're making Kiro our recommended AI-native development tool for Amazon...."

In October, Amazon revised its internal guidance for OpenAI's Codex to "Do Not Use" following a roughly six month assessment, according to a memo reviewed by Reuters. And Claude Code was briefly designated as "Do Not Use," before that was reversed following a reporter inquiry at the time.
The article adds that Amazon "has been fighting a reputation that it is trailing competitors in development of AI tools as rivals like OpenAI and Google speed ahead..."

"OpenAI is now internally testing 'ads' inside ChatGPT," reports BleepingComputer: Up until now, the ChatGPT experience has been completely free. While there are premium plans and models, you don't see GPT sell you products or show ads. On the other hand, Google Search has ads that influence your buying behaviour. OpenAI is planning to replicate a similar experience.

As spotted [by software engineer Tibor Blaho] on X.com,ChatGPT Android app 1.2025.329 beta includes new references to an "ads feature" with "bazaar content", "search ad" and "search ads carousel."

This move could disrupt the web economy, as what most people don't understand is that GPT likely knows more about users than Google. For example, OpenAI could create personalised ads on ChatGPT that promote products that you really want to buy... The leak suggests that ads will initially be limited to the search experience only, but this may change in the future.

An anonymous reader shared this report from CBS News: Artificial intelligence can do the work currently performed by nearly 12% of America's workforce, according to a recentstudy from the Massachusetts Institute of Technology. The researchers, relying on a metric called the "Iceberg Index" that measures a job's potential to be automated, conclude that AI already has the cognitive and technical capacity to handle a range of tasks in technology, finance, health care and professional services. The index simulated how more than 150 million U.S. workers across nearly 1,000 occupations interact and overlap with AI's abilities...

AI is also already doingsome of the entry-level jobsthat have historically been reserved for recent college graduates or relatively inexperienced workers, the report notes. "AI systems now generate more than a billion lines of code each day, prompting companies to restructure hiring pipelines and reduce demand for entry-level programmers," the researchers wrote. "These observable changes in technology occupations signal a broader reorganization of work that extends beyond software development."
"The study doesn't seek to shed light on how many workers AI may already have displaced or could supplant in the future," the article points out.

"To what extent such tools take over job functions performed by people depends on a number of factors, including individual businesses' strategy, societal acceptance and possible policy interventions, the researchers note."

"There are times when World of Tanks feels less like a videogame and more like a giant ad budget looking for something to be spent on," writes PC Gamer. This year, all those huge sacks with dollar signs on them have been thrown Benedict Cumberbatch's way, making him the game's newest "Holiday Ambassador" and the star of an absolutely bizarre Christmas advert. The story has very little to do with Christmas and, frankly, not much connection to tanks either, featuring Cumberbatch as a sort of chaotic, supernatural therapist trying to bring a meek nerd out of his shell with the help of a chaotic crowd of his other patients. It's a good watch, shedding the usual hard man action star vibe of past celebrity trailers in favour of something that feels more like a mischievous one act play.
Cumberbatch also portrayed Smaug and Sauron in The Hobbit films (2012-2014), Khan in Star Trek Into Darkness (2013), and Dr. Strange in six Marvel movies. And now Amazon has also hired Cumberbatch for what its calls its "Cannes-winning '5-Star Theater' campaign... performing real Amazon customer reviews as theatrical monologues." Cumberbatch performed over 15 reviews, including popular holiday gifts like the Bissell portable carpet cleaner, Toto bidet, and SharkNinja blender — showing that Amazon truly does have something for everyone on your list.
Last year Amazon produced a similar campaign starring Adam Driver ("Kylo Ren" from the final trilogy of Star Wars sequels). "The humor comes from the juxtaposition between Cumberbatch's gravitas and the text itself," reports Adweek, adding that the reviews were curated "using internal AI tools, to find the most oddly specific reviews on the platform."

Amazon will stream Cumberbatch's bizarre ads on major platforms including TikTok, Snapchat, YouTube, Lyft, Uber, Disney/Hulu, Paramount, and Roku, and on several NFL football games.

I remember when Amazon just chose the best funny fake reviews from customers, and then posted them on the front page of Amazon...

"The internet is being increasingly polluted by AI generated text, images and video," argues the site for a new browser extension called Slop Evader. It promises to use Google's search API "to only return content published before Nov 30th, 2022" — the day ChatGPT launched — "so you can be sure that it was written or produced by the human hand."

404 Media calls it "a scorched earth approach that virtually guarantees your searches will be slop-free." Slop Evader was created by artist and researcher Tega Brain, who says she was motivated by the growing dismay over the tech industry's unrelenting, aggressive rollout of so-called "generative AI" — despite widespread criticism and the wider public's distaste for it. "This sowing of mistrust in our relationship with media is a huge thing, a huge effect of this synthetic media moment we're in," Brain told 404 Media, describing how tools like Sora 2 have short-circuited our ability to determine reality within a sea of artificial online junk. "I've been thinking about ways to refuse it, and the simplest, dumbest way to do that is to only search before 2022...."

Currently, Slop Evader can be used to search pre-GPT archives of seven different sites where slop has become commonplace, including YouTube, Reddit, Stack Exchange, and the parenting site MumsNet. The obvious downside to this, from a user perspective, is that you won't be able to find anything time-sensitive or current — including this very website, which did not exist in 2022. The experience is simultaneously refreshing and harrowing, allowing you to browse freely without having to constantly question reality, but always knowing that this freedom will be forever locked in time — nostalgia for a human-centric world wide web that no longer exists.

Of course, the tool's limitations are part of its provocation. Brain says she has plans to add support for more sites, and release a new version that uses DuckDuckGo's search indexing instead of Google's. But the real goal, she says, is prompting people to question how they can collectively refuse the dystopian, inhuman version of the internet that Silicon Valley's AI-pushers have forced on us... With enough cultural pushback, Brain suggests, we could start to see alternative search engines like DuckDuckGo adding options to filter out search results suspected of having synthetic content (DuckDuckGo added the ability to filter out AI images in search earlier this year)... But no matter what form AI slop-refusal takes, it will need to be a group effort.

Earlier this month MasterCard noted that even Walmart now allows its customers to make purchases through ChatGPT. And after polling more than 4,000 consumers in the U.S., Canada, U.K., and UAE, they found "more than four in 10 consumers already use AI tools to help them shop, including 61% of Gen Z and 57% of millennials." Many (50% of Gen Z and 49% of millennials) say they'd even let AI handle all their gift-buying if it meant avoiding stress. Younger shoppers trust AI's taste, with 51% of Gen Z and 55% of millennials relying on it to deliver unique and thoughtful recommendations (sometimes even more than they trust themselves). The most popular uses include getting personalized product recommendations, confirming the best deal before purchasing, and summarizing thousands of reviews instantly. The bottom line: Shoppers are embracing AI as their new personal assistant — one that knows their budget, style, and patience level...

If the 2025 holiday shopper could be summed up in one word, it's intentional. They're planning earlier, spending wiser and using technology to make every dollar and every gift count.
The first figures are now in for the traditional "Black Friday" shopping day after Thanksgiving, and U.S. shoppers "spent a record $11.8 billion online," reports Reuters, "up 9.1% from 2024 on the year's biggest shopping day, according to Adobe Analytics, which tracks 1 trillion visits that shoppers make to online retail websites..."

And sure enough, this year shoppers were helped by AI: AI-powered shopping tools helped drive a surge in U.S. online spending on Black Friday, as shoppers bypassed crowded stores and turned to chatbots to compare prices and secure discounts amid concerns about tariff-driven price hikes... The AI-driven traffic to U.S. retail sites soared 805% compared to last year, Adobe said, when artificial intelligence tools such as Walmart's Sparky or Amazon's Rufus had not yet been launched. "Consumers are using new tools to get to what they need faster," said Suzy Davidkhanian, an analyst at eMarketer. "Gift giving can be stressful, and LLMs (large language models) make the discovery process feel quicker and more guided..." Globally, AI and agents influenced $14.2 billion in online sales on Black Friday, of which $3 billion came from the U.S. alone, according to software firm Salesforce.
There's another reason shoppers turned to AI. 2025's Black Friday arrived "amid tighter budgets, unemployment nearing a four-year high, U.S. consumer confidence sagging to a seven-month low and price tags that have shoppers watching every dollar," according to the article: Discount rates also remained flat when compared to 2024, with AI helping shoppers discover the best deals, and an increase in the price tags made deeper discounts difficult for retailers... Order volumes fell 1% as average selling prices rose 7%. Consumers also purchased fewer items at checkout, with units per transaction falling 2% on a year-over-year basis, Salesforce said.

The spending surge sets the stage for an even bigger Cyber Monday, projected to drive $14.2 billion in sales, up 6.3% on a year-over-year basis and the largest online shopping day of the year, Adobe said. Electronics are expected to see the deepest discounts on Cyber Monday, reaching 30% off list prices, along with strong deals on apparel and computers, Adobe said.

"By my count, Linux has over 11% of the desktop market," writes ZDNet's Steven Vaughan-Nichols: In StatCounter's latest US numbers, which cover through October, Linux shows up as only 3.49%. But if you look closer, "unknown" accounts for 4.21%. Allow me to make an educated guess here: I suspect those unknown desktops are actually running Linux. What else could it be? FreeBSD? Unix? OS/2? Unlikely. In addition, ChromeOS comes in at 3.67%, which strikes me as much too low. Leaving that aside, ChromeOS is a Linux variant. It just uses the Chrome web browser for its interface rather than KDE Plasma, Cinnamon, or another Linux desktop environment. Put all these together, and you get a Linux desktop market share of 11.37%...

If you want to look at the broader world of end-user operating systems, including phones and tablets, Linux comes out even better. In the US, where we love our Apple iPhones, Android — yes, another Linux distro — boasts 41.71% of the market share, according to StatCounter's latest numbers. Globally, however, Android rules with 72.55% of the market. Yes, that's right, if you widen the Linux end-user operating system metric to include PC, tablets, and smartphones, you can make a reasonable argument that Linux, and not Windows, is already the top dog operating system...

If you add Chrome OS (1.7%) and Android (15.8%), 23.3% of all people accessing the U.S. government's websites are Linux users. The Linux kernel's user-facing footprint is much larger than the "desktop Linux" label suggests.
The article lists reasons more people might be switching to Linux, including broader hardware support and "the increased viability of gaming via Steam and Proton" — but also the rise of Digital Sovereignty initiatives. (One EU group has even created EU OS.")

And finally, "not everyone is thrilled with Windows 11 being turned into an AI-agentic operating system."

An anonymous reader shared this report from CNBC: Italian defense company Leonardo on Thursday unveiled plans for an AI-powered shield for cities and critical infrastructure, adding to Europe's push to ramp up sovereign defense capabilities amid rising geopolitical tensions.

The system, dubbed the "Michelangelo Dome" in a nod to Israel's Iron Dome and U.S. President Donald Trump's plans for a "Golden Dome," will integrate multiple defense systems to detect and neutralize threats from sea to air including missile attacks and drone swarms... Leonardo's dome will be built on what CEO Roberto Cingolani called an "open architecture" system meaning it can operate alongside any country's defense systems... Leonardo's dome will be built on what CEO Roberto Cingolani called an "open architecture" system meaning it can operate alongside any country's defense systems.

Almost two-thirds of registered U.S. voters "say that a four-year college degree isn't worth the cost," according to a new NBC News poll: Just 33% agree a four-year college degree is "worth the cost because people have a better chance to get a good job and earn more money over their lifetime," while 63% agree more with the concept that it's "not worth the cost because people often graduate without specific job skills and with a large amount of debt to pay off." In 2017, U.S. adults surveyed were virtually split on the question — 49% said a degree was worth the cost and 47% said it wasn't. When CNBC asked the same question in 2013 as part of its All American Economic Survey, 53% said a degree was worth it and 40% said it was not. The eye-popping shift over the last 12 years comes against the backdrop of several major trends shaping the job market and the education world, from exploding college tuition prices to rapid changes in the modern economy — which seems once again poised for radical transformation alongside advances in AI...

Remarkably, less than half of voters with college degrees see those degrees as worth the cost: 46% now, down from 63% in 2013... The upshot is that interest in technical, vocational and two-year degree programs has soared.
"The 20-point decline over the last 12 years among those who say a degree is worth it — from 53% in 2013 to 33% now — is reflected across virtually every demographic group."

"A year after launching a commercial robotaxi service in Abu Dhabi, Chinese autonomous vehicle technology company WeRide and partner Uber can finally call that service driverless," reports TechCrunch.

A company official hailed it as "a historic transportation milestone, as the first driverless AV deployment outside of the U.S. or China." But TechCrunch notes that's just the beginning: Uber has spent the past two years locking up partnerships with 20 autonomous vehicle technology companies in various countries, including the United States, Europe, and the Middle East.

Those partnerships have expanded beyond the realm of robotaxis as well. Uber's deals span the full range of self-driving applications, including delivery and trucking. This year alone, it announced partnerships withAnn Arbor, Michigan-basedMay MobilityandVolkswagen, Chinese self-driving firms Momenta,Pony.ai, and Baidu, as well as a recent deal to create a premium robotaxi service using Lucid Gravity SUVs equipped with a self-driving system from San Francisco-based startup Nuro.

These deals are finally beginning to materialize into commercial services. For instance, Uber and Waymo launched a robotaxi service earlier this year in Austin. Now, Uber has expanded to the Middle East with WeRide in Abu Dhabi — with even more cities to come, including Dubai. Uber CEO Dara Khosrowshahi forecast in the company's third-quarter earnings report that there would be autonomous vehicle deployments on the Uber network in at least 10 cities by the end of 2026. Uber and WeRide have previously shared plans to expand to 15 cities throughout the Middle East and Europe, eventually scaling to thousands of robotaxis. That would represent a massive leap for WeRide, which today has more than 150 robotaxis in the region.

Digital Trends reports: A wave of shortages now threatens to ripple across RAM, SSDs, and even hard drives, affecting not only performance-hungry rigs but also everyday systems.

— CyberPowerPC has publicly confirmed it will raise prices on all systems starting December 7th due to RAM costs spiking by 500% and SSD prices doubling since October.

— Memory suppliers warn of a global DRAM and SSD shortage running into late 2026 or even 2027, driven heavily by AI server demand.

— As reported by Bloomberg, Lenovo has already stockpiled memory to ride out the crunch and maintain steadier PC pricing.

— Among other OEMs, HP, in its recent earnings call, flagged possible price increases or lower-spec models on the back of rising component costs.
But Apple "may also be in a good position to weather the shortage," reports Ars Technica, since "analysts at Morgan Stanley and Bernstein Research believe that Apple has already laid claim to the RAM that it needs and that its healthy profit margins will allow it to absorb the increases better than most."

Ars Technica also shows how much RAM and storage prices have jumped — sometimes as much as 2x or even 3x in just three months. "In short, there's no escaping these price increases, which affect SSDs and both DDR4 and DDR5 RAM kits of all capacities (though higher-capacity RAM kits do seem to be hit a little harder)." Memory and storage shortages can be particularly difficult to get through. As with all chips, it can take years to ramp up capacity and/or build new manufacturing facilities... And memory makers in particular may be slow to ramp up manufacturing capacity in response to shortages. If they decide to start manufacturing more chips now, what happens if memory demand drops off a cliff in six months or a year (if, say, an AI bubble deflates or pops altogether)? It means an oversupply of memory chips — consumers benefit from rock-bottom prices for components, but it becomes harder for manufacturers to cover their costs... The upshot is: Not only are memory prices getting bad now, but it's exceptionally difficult to predict when shortage-fueled price hikes might end...

Tom's Hardware reports that AMD has told its partners that it expects to raise GPU prices by about 10 percent starting next year and that Nvidia may have canceled a planned RTX 50-series Super launch entirely because of shortages and price increases.

An EDM song by the British group Haven ran into trouble in October after it shared clips of upcoming song "I Run" on TikTok.

The song "was an overnight viral sensation online," writes Digital Music News — racking up millions of plays "even before it hit streaming services." (Although the Washington Post notes that "Record labels and TikTok users began questioning whether 'I Run' used an AI deepfake, modeled off British R&B singer Jorja Smith, for the vocals.")

Digital Music News picks up the story: The artist says he used his own voice to record the vocals, and then ran it through layers of processing and filtering to turn it into the female-sounding voice heard in the track. However, that filtering also included the use of the controversial genAI platform Suno — and that's what complicates things... [The article says later that Suno "is currently in the middle of a blockbuster lawsuit with the Big Three major labels over allegations of widespread copyright infringement of sound recordings used during the AI model training process."]

Meanwhile, the song was rapidly amassing listenership. It soared to #11 on the U.S. Spotify chart and #25 on Spotify globally. Videos using the song continued going viral on TikTok and Instagram, including one in which rapper Offset had apparently played the song during a Boiler Room set, which later turned out to be falsified. And then, as quickly as it appeared, "I Run" was taken down from streaming services, including Spotify and Apple Music. That was due, in part, to numerous takedown notices from The Orchard, the label to which Jorja Smith is signed, as well as the RIAA and IFPI. The takedown notices alleged various issues with the track, including the "misrepresentation" of another artist, as well as copyright infringement.

As a result, the song has also been withheld from the Billboard charts, including the Hot 100, on which it had been predicted to debut this week before the controversy. Billboard points out that it "reserves the right to withhold or remove titles from appearing on the charts that are known to be involved in active legal disputes related to copyright infringement that may extend to the deletion of such content on digital service providers."
The song itself has now been re-released with an all-human vocal track. But going forward will the music industry ever work with AI platforms? The Washington Post reports: "I Run" has taken off as record labels remain unsure of the extent to which they should welcome generative AI programs such as Suno or Udio into the industry. After the two AI music companies began growing in popularity, the three major labels — Sony Music, Warner Music Group and Universal Music Group — filed lawsuits against Suno and Udio, claiming that the AI companies have used the labels' sound recordings to train their model.

Since then, UMG and Warnerhave reached agreementsto work with Udio, ending their litigation... It comes shortly after all three major labels licensed their catalogue to Klay, a music streaming start-up that allows users to adjust songs using artificial intelligence. Major licensing organizations such as ASCAP and BMI shared that they would register songs that were partially AI-generated — but not fully generated ones.

Haven appears to present an uncomfortable edge case. While some AI-generated songs that sound broadly like other artists have been allowed to remain on streaming platforms, the voice in "I Run" appears to have been deemed too duplicative for comfort.

OpenAI's data centre partners are on course to amass almost $100 billion in borrowing tied to the lossmaking start-up, as the ChatGPT maker benefits from a debt-fuelled spending spree without taking on financial risks itself. Financial Times: SoftBank, Oracle and CoreWeave have borrowed at least $30 billion to invest in the start-up or help build its data centres, according to FT analysis. Investment group Blue Owl Capital and computing infrastructure companies such as Crusoe also rely on deals with OpenAI to service about $28 billion in loans.

A group of banks is in talks to lend another $38 billion for Oracle and data centre builder Vantage to fund further sites for OpenAI, according to people familiar with the matter. The deal is expected to be finalised in the coming weeks. OpenAI executives have said they plan to raise substantial debt to help pay for these contracts, but so far the financial burden has fallen to its counterparties and their lenders. "That's been kind of the strategy," said a senior OpenAI executive. "How does [OpenAI] leverage other people's balance sheets?"

Singapore has claimed the top spot in the 2025 Global Talent Competitiveness Index for the first time, displacing Switzerland from a position the European nation had held since the ranking's inception in 2013. The index, produced by business school INSEAD and the Portulans Institute, measured 135 economies across 77 indicators spanning soft skills, AI talent concentration, and formal education systems. The city-state ranked first globally in formal education and what the report calls "Generalist Adaptive Skills," a category covering soft skills, digital literacy, and innovation-oriented thinking.

A key factor in Singapore's rise was a seven-place jump in talent retention, moving from 38th to 31st. The United States fell from third place in 2023 to ninth this year, its weakest showing in 12 years, due to declines in openness and lifelong learning metrics. High-income European countries continue to dominate the top ten, holding seven positions.

China installed 295,000 industrial robots last year -- nearly nine times as many as the United States and more than the rest of the world combined -- as the country races to automate its manufacturing base amid rising labor costs at home and tariff threats from abroad.

The nation's stock of operational robots surpassed 2 million in 2024, according to the International Federation of Robotics. Of 131 factories globally recognized by the World Economic Forum for boosting productivity through cutting-edge technologies like AI, 45 are in mainland China compared to three in the US.

At Midea's washing machine factory in Jingzhou, an AI "factory brain" manages 14 virtual agents that coordinate robots and machines on the floor. The home-appliance giant reports that its revenue per employee grew nearly 40% between 2015 and 2024, and processes that once took 15 minutes now take 30 seconds. Down jacket maker Bosideng has cut sample production time from 100 days to 27 days using AI design tools, reducing development costs by 60%. At the port of Tianjin, scheduling that previously required 24 hours now takes 10 minutes, and 88% of large container equipment is automated. The port's operator says it requires 60% fewer workers than traditional facilities.

An analysis of submissions to next year's International Conference on Learning Representations has found that roughly one in five peer reviews were fully generated by AI, a discovery that came after researchers including Carnegie Mellon's Graham Neubig grew suspicious of feedback on their manuscripts that seemed unusually verbose and requested non-standard statistical analyses.

Neubig posted on X offering a reward for anyone who could scan the conference's submissions for AI-generated text, and Max Spero, CEO of detection tool developer Pangram Labs, responded the next day. Pangram screened all 19,490 studies and 75,800 peer reviews submitted to ICLR 2026, finding that 21% of reviews were fully AI-generated and more than half showed signs of AI use. The conference had permitted AI tools for polishing text but prohibited falsified content. Each reviewer was assigned five papers to review in two weeks on average -- a load that senior programme chair Bharath Hariharan described as "much higher than what has been done in the past."

ChatGPT can browse the web, write code and analyze images, but ask it what time it is and you might get the correct answer, a confident wrong answer, or a polite refusal -- sometimes all three within minutes of each other.

The problem stems from how large language models work. These systems predict answers based on training data and don't receive constant real-time updates about things like time unless they specifically search the internet. AI robotics expert Yervant Kulbashian told The Verge that a language model "is only referencing things that have entered this space," comparing it to a castaway on an island stocked with books but no watch.

OpenAI can give ChatGPT access to system clocks, and does so through features like Search. But there are tradeoffs: every clock check consumes space in the model's context window, the finite portion of information it can hold at any given moment. Pasquale Minervini, a natural language processing researcher at the University of Edinburgh, said the leading models also struggle to read analog clock faces and have trouble with calendars.

The U.S. Patent and Trademark Office has issued new guidelines outlining when inventions created with the help of AI can be patented. From a report: USPTO Director John Squires said on Wednesday in a notice set to be published Friday, that the office considers generative AI systems to be "analogous to laboratory equipment, computer software, research databases, or any other tool that assists in the inventive process."

"They may provide services and generate ideas, but they remain tools used by the human inventor who conceived the claimed invention," the office said. "When one natural person is involved in creating an invention with the assistance of AI, the inquiry is whether that person conceived the invention under the traditional conception standard."

The office reiterated its guidance from last year that AI itself cannot be considered an inventor under U.S. patent law. However, it rejected the approach taken by the PTO during former President Joe Biden's administration for deciding when AI-assisted inventions are patentable, which relied on a standard normally used to determine when multiple people can qualify as joint inventors.

The CEO of Epic Games, Tim Sweeney, has argued that platforms like Steam should not label games that are made using AI. From a report: Responding to a post on Twitter from a user who suggested that storefronts drop this tag, the industry exec said that it "makes no sense" to flag such content. Sweeney added that soon AI will be a part of the way all games are made. "The AI tag is relevant to art exhibits for authorship disclosure, and to digital content licensing marketplaces where buyers need to understand the rights situation," Sweeney said. "It makes no sense for game stores, where AI will be involved in nearly all future production."

Researchers at MIT and Oak Ridge National Laboratory have built a simulation that models all 151 million American workers and their skills, then maps those skills against the capabilities of over 13,000 AI tools currently in production to see where the two overlap. The answer, according to their analysis: 11.7% of the US labor market's total wage value, or about $1.2 trillion, sits in tasks that AI systems can technically perform [PDF].

The researchers call this the Iceberg Index, and the name is deliberate. The visible AI disruption happening in tech jobs right now accounts for only 2.2% of labor market wage value. The remaining exposure lurks in cognitive and administrative work across finance, healthcare administration, and professional services, and unlike tech-sector disruption, it's spread across all fifty states rather than concentrated on the coasts.

Delaware and South Dakota show higher Iceberg Index values than California because their economies lean heavily on administrative and financial work. Ohio and Tennessee register modest tech-sector exposure but substantial hidden risk in the white-collar functions that support their manufacturing bases.

To validate the framework, the researchers compared their predictions against Anthropic's Economic Index tracking real-world AI usage from millions of Claude users. The two measures agreed on state categorizations 69% of the time, with particularly strong alignment at the extremes.

The Iceberg Index doesn't predict job losses or adoption timelines. It measures technical capability, the overlap between what AI can do and what occupations require. Traditional economic indicators like GDP and unemployment explain less than five percent of the variation in this skill-based exposure, which is partly why the researchers argue workforce planners need new metrics.