The U.S. may have led China in the AI race for the past decade, according to Alexandr Wang, CEO of Scale AI, but on Christmas Day, everything changed. From a report: Wang, whose company provides training data to key AI players including OpenAI, Google and Meta , said Thursday at the World Economic Forum in Davos, Switzerland, that DeepSeek, the leading Chinese AI lab, released an "earth-shattering model" on Christmas Day, then followed it up with a powerful reasoning-focused AI model, DeepSeek-R1, which competes with OpenAI's recently released o1 model.

"What we've found is that DeepSeek ... is the top performing, or roughly on par with the best American models," Wang said. In an interview with CNBC, Wang described the artificial intelligence race between the U.S. and China as an "AI war," adding that he believes China has significantly more Nvidia H100 GPUs -- AI chips that are widely used to build leading powerful AI models -- than people may think, especially considering U.S. export controls. [...] "The United States is going to need a huge amount of computational capacity, a huge amount of infrastructure," Wang said, later adding, "We need to unleash U.S. energy to enable this AI boom." DeepSeek's holding company is a quant firm, which happened to have a lot of GPUs for trading and mining. DeepSeek is their "side project."

President Donald Trump signed an executive order on AI Thursday that will revoke past government policies his order says "act as barriers to American AI innovation." From a report: To maintain global leadership in AI technology, "we must develop AI systems that are free from ideological bias or engineered social agendas," Trump's order says. The new order doesn't name which existing policies are hindering AI development but sets out to track down and review "all policies, directives, regulations, orders, and other actions taken" as a result of former President Joe Biden's sweeping AI executive order of 2023, which Trump rescinded Monday.

Any of those Biden-era actions must be suspended if they don't fit Trump's new directive that AI should "promote human flourishing, economic competitiveness, and national security." Last year, the Biden administration issued a policy directive that said U.S. federal agencies must show their artificial intelligence tools aren't harming the public, or stop using them. Trump's order directs the White House to revise and reissue those directives, which affect how agencies acquire AI tools and use them.

An anonymous reader quotes a report from The Verge: TerraPower, a nuclear energy startup founded by Bill Gates, struck a deal this week with one of the largest data center developers in the US to deploy advanced nuclear reactors. TerraPower and Sabey Data Centers (SDC) are working together on a plan to run existing and future facilities on nuclear energy from small reactors. A memorandum of understanding signed by the two companies establishes a "strategic collaboration" that'll initially look into the potential for new nuclear power plants in Texas and the Rocky Mountain region that would power SDC's data centers. [...]

There's still a long road ahead before that can become a reality. The technology TerraPower and similar nuclear energy startups are developing still have to make it through regulatory hurdles and prove that they can be commercially viable. Compared to older, larger nuclear power plants, the next generation of reactors are supposed to be smaller and easier to site. Nuclear energy is seen as an alternative to fossil fuels that are causing climate change. But it still faces opposition from some advocates concerned about the impact of uranium mining and storing radioactive waste near communities. TerraPower's reactor design for this collaboration, Natrium, is the only advanced technology of its kind with a construction permit application for a commercial reactor pending with the U.S. Nuclear Regulatory Commission, according to the company. The company just broke ground on a demonstration project in Wyoming last year, and expects it to come online in 2030.

The World Economic Forum's Global Risks Report 2025 (PDF) highlights misinformation as the top global risk due to generative AI tools and state-sponsored campaigns undermining democratic systems, while cyberespionage ranks as a persistent threat with inadequate cyber resilience, especially among small organizations. From a report: The manipulation of information through gen AI and state-sponsored campaigns is disrupting democratic systems and undermining public trust in critical institutions. Efforts to combat this risk have a "formidable opponent" in gen AI-created false or misleading content that can be produced and distributed at scale, the report warned. Misinformation campaigns in the form of deepfakes, synthetic voice recordings or fabricated news stories are now a leading mechanism for foreign entities to influence "voter intentions, sow doubt among the general public about what is happening in conflict zones, or tarnish the image of products or services from another country." This is especially acute in India, Germany, Brazil and the United States.

Concern remains especially high following a year of the so-called "super elections," which saw heightened state-sponsored campaigns designed to manipulate public opinion. But while it has become increasingly difficult to distinguish AI-generated fake content from human-generated one, AI technologies, in itself, is low in WEF's risk ranking. In fact, it has declined in the two-year outlook, from 29 in last year's report to 31 this year.

Cyberespionage and warfare continue to be a reason for unease for most organizations, ranked fifth in the global risk landscape. According to the report, one in three CEOs cited cyberespionage and intellectual property theft as their top concerns in 2024. Seventy-one percent of chief risk officers say cyber risk and criminal activity such as money laundering and cybercrime could severely impact their organizations, while 45% of cyber leaders are concerned about disruption of operations and business processes, according to WEF's Global Cybersecurity Outlook 2025 report. The rising likelihood of threat actor activity and sophisticated technological disruption is listed as immediate concerns among security leaders.

An anonymous reader quotes a report from 404 Media: A pseudonymous coder has created and released an open source "tar pit" to indefinitely trap AI training web crawlers in an infinitely, randomly-generating series of pages to waste their time and computing power. The program, called Nepenthes after the genus of carnivorous pitcher plants which trap and consume their prey, can be deployed by webpage owners to protect their own content from being scraped or can be deployed "offensively" as a honeypot trap to waste AI companies' resources.

"It's less like flypaper and more an infinite maze holding a minotaur, except the crawler is the minotaur that cannot get out. The typical web crawler doesn't appear to have a lot of logic. It downloads a URL, and if it sees links to other URLs, it downloads those too. Nepenthes generates random links that always point back to itself -- the crawler downloads those new links. Nepenthes happily just returns more and more lists of links pointing back to itself," Aaron B, the creator of Nepenthes, told 404 Media. "Of course, these crawlers are massively scaled, and are downloading links from large swathes of the internet at any given time," they added. "But they are still consuming resources, spinning around doing nothing helpful, unless they find a way to detect that they are stuck in this loop." You can try Nepenthes via this link (it loads slowly and links endlessly on purpose).

The rumors are true: OpenAI today launched Operator, an AI agent capable of performing web-based tasks through its own browser, as a research preview for U.S. subscribers of its $200 monthly ChatGPT Pro tier. The agent uses GPT-4's vision capabilities and reinforcement learning to interact with websites through mouse and keyboard actions without requiring API integration, OpenAI said in a blog post.

Operator can self-correct and defer to users for sensitive information though there are some limitations with complex interfaces. OpenAI said it's partnering with DoorDash, Instacart, OpenTable and others to develop real-world applications, with plans to expand access to Plus, Team and Enterprise users.

Check out our list of the best AI web browsing agents.

Bruce Schneier and Nathan E. Sanders, writing in a post: Someone who makes calculus mistakes is also likely to respond "I don't know" to calculus-related questions. To the extent that AI systems make these human-like mistakes, we can bring all of our mistake-correcting systems to bear on their output. But the current crop of AI models -- particularly LLMs -- make mistakes differently.

AI errors come at seemingly random times, without any clustering around particular topics. LLM mistakes tend to be more evenly distributed through the knowledge space. A model might be equally likely to make a mistake on a calculus question as it is to propose that cabbages eat goats. And AI mistakes aren't accompanied by ignorance. A LLM will be just as confident when saying something completely wrong -- and obviously so, to a human -- as it will be when saying something true. The seemingly random inconsistency of LLMs makes it hard to trust their reasoning in complex, multi-step problems. If you want to use an AI model to help with a business problem, it's not enough to see that it understands what factors make a product profitable; you need to be sure it won't forget what money is.

[...] Humans may occasionally make seemingly random, incomprehensible, and inconsistent mistakes, but such occurrences are rare and often indicative of more serious problems. We also tend not to put people exhibiting these behaviors in decision-making positions. Likewise, we should confine AI decision-making systems to applications that suit their actual abilities -- while keeping the potential ramifications of their mistakes firmly in mind.

Microsoft's absence from OpenAI's Stargate announcement follows months of tension between the companies and signals a new era in which the longtime partners will be less reliant on each other. From a report: At a White House press conference, the ChatGPT maker announced Stargate, a venture with Oracle and tech investor SoftBank. The new company plans to spend up to $500 billion building new data centers in the U.S. to help power OpenAI's development.

The assembled leaders -- OpenAI's Sam Altman, Oracle's Larry Ellison, SoftBank's Masayoshi Son and President Trump -- discussed how AI could create jobs and even cure cancer. Microsoft CEO Satya Nadella was thousands of miles away, at the World Economic Forum in Davos, Switzerland. The developments show how the OpenAI-Microsoft partnership that helped trigger the generative-AI boom is drifting apart as each company focuses on its own evolving needs.

In the months leading up to the announcement, the two sides had been haggling over what to do about OpenAI's seemingly insatiable appetite for computing power and its contention Microsoft couldn't fulfill it even though their agreement didn't allow OpenAI to easily switch to others, said people familiar with the discussions. OpenAI is almost entirely reliant on Microsoft to provide it with the data centers it needs to build and operate its sophisticated AI software. That has been part of their agreement since Microsoft first invested in 2019. With the success of ChatGPT, OpenAI's need for computing power surged. Its executives have said ending the exclusive cloud contract could be crucial to compete with rival AI developers that don't have the same constraints.

Consumer spending on apps is projected to reach $150 billion globally in 2024, up 13% from the prior year. According to Sensor Tower's annual "State of Mobile" report, it's being fueled by a 200% surge in spending on generative AI apps like ChatGPT and Gemini, which collectively drew $1.1 billion. TechCrunch reports: If this rate of growth is sustained, this category of apps could move into the top 10 by consumer spending within a year, the firm notes. Though the release of new AI models, like OpenAI's GPT-4o last summer, helped drive app revenue up to record numbers at times, consumer demand for AI apps was consistent throughout the year -- not only during these peak surges. As a result, consumers spent nearly 7.7 billion hours using AI apps in 2024, while apps mentioning "AI" were also downloaded 17 billion times in the year.

ChatGPT alone reached 50 million monthly active users -- faster than Temu, Disney+, or YouTube Music, for comparison. This indicates there's still a growing appetite for AI apps and those with AI features.

LinkedIn has been sued by Premium customers alleging the platform disclosed private messages to third parties without consent to train generative AI models. The lawsuit seeks damages for breach of contract and privacy violations, accusing LinkedIn of attempting to minimize scrutiny over its actions. Reuters reports: According to a proposed class action filed on Tuesday night on behalf of millions of LinkedIn Premium customers, LinkedIn quietly introduced a privacy setting last August that let users enable or disable the sharing of their personal data. Customers said LinkedIn then discreetly updated its privacy policy on Sept. 18 to say data could be used to train AI models, and in a "frequently asked questions" hyperlink said opting out "does not affect training that has already taken place."

This attempt to "cover its tracks" suggests LinkedIn was fully aware it violated customers' privacy and its promise to use personal data only to support and improve its platform, in order to minimize public scrutiny and legal fallout, the complaint said. The lawsuit was filed in the San Jose, California, federal court on behalf of LinkedIn Premium customers who sent or received InMail messages, and whose private information was disclosed to third parties for AI training before Sept. 18. It seeks unspecified damages for breach of contract and violations of California's unfair competition law, and $1,000 per person for violations of the federal Stored Communications Act. LinkedIn said in a statement: "These are false claims with no merit."

In the aftermath of Israel's October 2023 ground invasion of Gaza, Google reportedly worked with the Israeli military to provide AI services while racing against Amazon for contracts. This comes despite publicly denying collaboration with the military and punishing employees protesting its involvement in Project Nimbus, a $1.2 billion cloud computing agreement with Israel. The Verge reports: In the weeks after Hamas's October 7th attack on Israel, employees at Google's cloud division worked directly with the Israel Defense Forces (IDF) -- even as the company told both the public and its own employees that Google only worked with civilian government ministries, the documents reportedly show.

Weeks after the war began, an employee with Google's cloud division escalated the IDF's military's requests for access to Google's AI technology, according to the Washington Post. In another document, an employee warned that Google needed to quickly respond to the military's requests, or else Israel would turn to Amazon for its cloud computing needs. In a November 2023 document, an employee thanks a coworker for handling the IDF's request. Months later, employees requested additional access to AI tools for the IDF.

At Galaxy Unpacked today in San Jose, California, Samsung unveiled the new Galaxy S25 series of flagship smartphones loaded with AI capabilities and LLMs. "Currently, the Galaxy S25 range is comprised of the Galaxy S25 ($800), Galaxy S25+ ($1,000), and Galaxy S25 Ultra ($1,300)," reports Wired. "The phones are available for preorder today and will officially go on sale February 7." Since the hardware is relatively unchanged from last year's Galaxy S24 series, here's what Wired has to say about the new AI smarts: The Galaxy S25 is a tale of two AIs: Gemini and Bixby. Yes, while Google's Gemini AI assistant sits at the forefront -- it can finally be triggered through a long press of the power button-- Samsung is bringing its original Bixby voice assistant out from the shadows. Bixby has been enhanced with large language models but is still designed to handle phone functions, like changing device settings. Gemini is meant to be used for general web queries and more complex actions. You can even have two hot words, one for each assistant. I foresee all of this being confusing [...].

The highlight AI feature debuting on the Galaxy S25 series is "cross-app experiences." These are tasks you can ask Gemini to perform, even if the task requires multiple apps. For example, you can ask for the schedule of this season's Arsenal matches and then add it to your calendar; Gemini will then search and add every Arsenal FC game in the season to your schedule. Or you can ask it to find pet-friendly vegan restaurants nearby and text the list to a friend. It even works with images too -- snap a pic of your fridge and ask Gemini to find you a recipe based on the available ingredients. These cross-app experiences work with Google apps, Samsung's Galaxy apps, and select third-party apps, like WhatsApp and Spotify.

All these AI features have culminated in a new app: Now Brief. Samsung calls this proactive assistance (remember Google's Now on Tap?) where a morning brief arrives with the weather, upcoming calendar events, stock details, news articles, and suggestions to trigger routines. There's also an evening brief with a summary of the day's events with photos. Since the feature can plug into email, it'll send reminders about expiring coupons and upcoming travel tickets. Samsung claims it can even suggest changing an 8:45 am alarm even earlier if it sees a 9 am meeting on the schedule. On the lock screen, a "Now Bar" widget persists at the bottom, much like Apple's Live Activities. It'll offer quick access to the Now Brief app, but it will also show updates for favorite sports teams, along with glanceable directions from Google Maps.

The rest of the AI features are playing a bit of catch-up to Apple and Google's Pixel phones. There's Drawing Assist, a generative AI tool to craft new images in different art styles based on sketches or text prompts. AI Select works with the S Pen stylus on the S25 Ultra and understands what is selected -- for example, if a video is selected, it will suggest turning it into a GIF. Audio Eraser is an editing tool to cut out background noise in videos post-capture, canceling out the sound of a crowd's chatter or an ambulance's siren. Finally, Samsung's Generative Edit feature, which lets you erase unwanted objects in images, now works locally on the device and is much more accurate and faster. A full list of specs can be found here. You can watch a recording of the event on YouTube.

An anonymous reader quotes a report from Ars Technica: On Tuesday, Anthropic CEO Dario Amodei predicted that AI models may surpass human capabilities "in almost everything" within two to three years, according to a Wall Street Journal interview at the World Economic Forum in Davos, Switzerland. Speaking at Journal House in Davos, Amodei said, "I don't know exactly when it'll come, I don't know if it'll be 2027. I think it's plausible it could be longer than that. I don't think it will be a whole bunch longer than that when AI systems are better than humans at almost everything. Better than almost all humans at almost everything. And then eventually better than all humans at everything, even robotics."

Amodei co-founded Anthropic in 2021 with his sister Daniela Amodei and five other former OpenAI employees. Not long after, Anthropic emerged as a strong technological competitor to OpenAI's AI products (such as GPT-4 and ChatGPT). Most recently, its Claude 3.5 Sonnet model has remained highly regarded among some AI users and highly ranked among AI benchmarks. During the WSJ interview, Amodei also spoke some about the potential implications of highly intelligent AI systems when these AI models can control advanced robotics. "[If] we make good enough AI systems, they'll enable us to make better robots. And so when that happens, we will need to have a conversation... at places like this event, about how do we organize our economy, right? How do humans find meaning?"

He then shared his concerns about how human-level AI models and robotics that are capable of replacing all human labor may require a complete re-think of how humans value both labor and themselves. "We've recognized that we've reached the point as a technological civilization where the idea, there's huge abundance and huge economic value, but the idea that the way to distribute that value is for humans to produce economic labor, and this is where they feel their sense of self worth," he added. "Once that idea gets invalidated, we're all going to have to sit down and figure it out." The eye-catching comments, similar to comments about AGI made recently by OpenAI CEO Sam Altman, come as Anthropic negotiates a $2 billion funding round that would value the company at $60 billion. Amodei disclosed that Anthropic's revenue multiplied tenfold in 2024. Further reading: Salesforce Chief Predicts Today's CEOs Will Be the Last With All-Human Workforces

Salesforce CEO Marc Benioff said Wednesday that current business leaders may be the last generation to manage an exclusively human workforce, as AI transforms the workplace. "We are really moving into a world now of managing humans and agents together," he told Axios.

His company's Agentforce platform, launched in September, has taken over many customer support tasks, prompting plans to move support staff into sales roles. Speaking to Axios at the World Economic Forum, Benioff dismissed Microsoft's AI CoPilot as disappointing and promised to defend his employees against discrimination amid political debates over corporate diversity programs.

Santee Cooper, the big power provider in South Carolina, has tapped financial advisers to look for buyers that can restart construction on a pair of nuclear reactors that were mothballed years ago. From a report: The state-owned utility is betting interest will be strong, with tech giants such as Amazon.com and Microsoft in need of clean energy to fuel data centers for artificial-intelligence capabilities. The details Santee Cooper announced Wednesday it is seeking proposals for buyers to complete the project at South Carolina's sprawling V.C. Summer Nuclear Station, confirming an earlier report from The Wall Street Journal.

The utility is working with bankers at Centerview Partners, which will accept proposals until May 5. Santee Cooper will likely look to tap a consortium that could include a construction firm, a tech company that will use the power and an additional partner for capital, according to people familiar with the matter. It is also looking for another power company partner because it doesn't plan to own or operate the units once they are up and running.

Search in Premiere Pro has been updated with AI-powered visual recognition, allowing users to find videos by describing the contents of the footage. From a report: It's just one of several quality-of-life features Adobe is adding to Premiere Pro, After Effects, and Frame.io that aim to save video editors time on their projects. Users can enter search terms like "a person skating with a lens flare" to find corresponding clips within their media library.

Adobe says the media intelligence AI can automatically recognize "objects, locations, camera angles, and more," alongside spoken words -- providing there's a transcript attached to the video. The feature doesn't detect audio or identify specific people, but it can scrub through any metadata attached to video files, which allows it to fetch clips based on shoot dates, locations, and camera types. The media analysis runs on-device, so doesn't require an internet connection, and Adobe reiterates that users' video content isn't used to train any AI models.

Microsoft will pay to restore parts of Brazil's Amazon and Atlantic forests [non-paywalled source] in exchange for hundreds of millions of dollars' worth of carbon credits, becoming the latest Big Tech player to bet that nature-based solutions can offset an artificial intelligence-driven surge in greenhouse gas emissions. Financial Times: The $3.2tn US company told the Financial Times it had signed a deal to buy 3.5mn credits over 25 years from Re.green, a Brazilian start-up which buys up farming and cattle land. It restores the land by planting native tree species, in projects financed through carbon credits and timber sales. Neither company disclosed a value for the deal, but recent market analysis suggests it could be worth around $200mn.

Microsoft's recent dealmaking has made it one of the biggest buyers of nature-based carbon removals globally. The deal comes as groups including Microsoft, Google and Amazon invest heavily in data centres to cope with the huge demand stemming from the growth of generative AI. But the buildout is leading to a surge in their energy usage and complicating their pledges to investors to curb emissions.

Alphabet's Google is backing AI developer Anthropic with a further $1 billion, building its stake in one of the most promising rivals to OpenAI. From a report: The new funding comes in addition to more than $2 billion that Google has already invested in Anthropic, according to a person familiar with the deal, who asked not to be named discussing a private matter. Google has a business agreement with Anthropic that covers the use of a suite of online tools and services.

Amazon counts among its biggest backers. San Francisco-based Anthropic is best known for its Claude family of large language models, which compete with OpenAI's GPT. Like its peers, the company has been raising significant sums to sustain investment in expanding its computing capabilities and keep pace in a race to advance AI. The new deal comes weeks after Bloomberg News reported that Anthropic is in advanced talks to raise $2 billion in a funding round led by Lightspeed Venture Partners that would value the startup at $60 billion.

Microsoft, the biggest investor in OpenAI and its principal cloud partner, is losing its designation as exclusive provider of computing capacity for the artificial intelligence startup. CNBC: In a blog post on Tuesday, Microsoft said that it's still in a favorable position with OpenAI. Going forward, when OpenAI seeks additional capacity, Microsoft will have the "right of first refusal" before OpenAI checks with other parties. The change in their relationship was disclosed as part of President Donald Trump's announcement of the Stargate Project, a joint venture with OpenAI, Oracle and Softbank to invest billions of dollars in AI infrastructure in the U.S.

Executives from those companies committed to invest an initial $100 billion and up to $500 billion over the next four years in the project, which will be set up as a separate company. Oracle is a "key initial technology partner" alongside Arm, Microsoft and Nvidia in setting up data center infrastructure, OpenAI said in a blog post. JPMorgan, in a note to clients: My takeaway is that MSFT is somewhat reading the room on capex. Softbank and Oracle are taking on some of the financial burden (in order to get some skin in the game) while MSFT still maintains access to OpenAI IP through to 2030 and has right of first refusal on any OpenAI new capacity. Feels like a good outcome for MSFT to me.

Apple's upcoming updates -- macOS Sequoia 15.3, iOS 18.3, and iPadOS 18.3 -- will enable Apple Intelligence by default on compatible devices, requiring users to manually disable it if undesired. From Apple's developer release notes: "For users new or upgrading to iOS 18.3, Apple Intelligence will be enabled automatically during iPhone onboarding. Users will have access to Apple Intelligence features after setting up their devices. To disable Apple Intelligence, users will need to navigate to the Apple Intelligence & Siri Settings pane and turn off the Apple Intelligence toggle. This will disable Apple Intelligence features on their device." MacRumors reports: With macOS Sequoia 15.1, macOS Sequoia 15.2, iOS 18.1, and iOS 18.2, Apple Intelligence was opt-in rather than opt-out, and users who wanted the feature needed to turn it on in the Settings app. Going forward, it will be enabled by default, and Mac, iPhone, and iPad users who do not want to use the feature will need to turn it off. The report notes that macOS Sequoia 15.3 introduces Genmoji, allowing Mac users to create custom emoji characters, and enhances Notification summaries with clearer indicators for AI-generated information.

Public releases of this and other software updates are expected next week, following today's release candidate versions.