A new paper [PDF] from the AI Disclosures Project claims OpenAI likely trained its GPT-4o model on paywalled O'Reilly Media books without a licensing agreement. The nonprofit organization, co-founded by O'Reilly Media CEO Tim O'Reilly himself, used a method called DE-COP to detect copyrighted content in language model training data.

Researchers analyzed 13,962 paragraph excerpts from 34 O'Reilly books, finding that GPT-4o "recognized" significantly more paywalled content than older models like GPT-3.5 Turbo. The technique, also known as a "membership inference attack," tests whether a model can reliably distinguish human-authored texts from paraphrased versions.

"GPT-4o [likely] recognizes, and so has prior knowledge of, many non-public O'Reilly books published prior to its training cutoff date," wrote the co-authors, which include O'Reilly, economist Ilan Strauss, and AI researcher Sruly Rosenblat.

With the signing of HB25-1040 on Monday, Colorado now defines nuclear as a "clean energy resource" since it doesn't release large amounts of climate-warming emissions. "The category was previously reserved for renewables like wind, solar and geothermal, which don't carry the radioactive stigma that's hobbled fission power plants following disasters like Chernobyl and Fukushima," notes Colorado Public Radio. From the report: In an emailed statement, Ally Sullivan, a spokesperson for the governor's office, said the law doesn't advance any specific nuclear energy project, and no utility has proposed building a nuclear power plant in Colorado. It does, however, allow nuclear energy to potentially serve as one piece of the state's plan to tackle climate change. "If nuclear energy becomes sufficiently cost-competitive, it could potentially become part of Colorado's clean energy future. However, it must be conducted safely, without harming communities, depleting other natural resources or replacing other clean energy sources," Sullivan said.

By redefining nuclear energy as "clean," the law would let future fission-based power plants obtain local grants previously reserved for other carbon-free energy sources, and it would allow those projects to contribute to Colorado's renewable energy goals. It also aligns state law with a push to reshape public opinion of nuclear energy. Nuclear energy proponents promise new reactor designs are smaller and safer than hulking power plants built in the 20th century. By embracing those systems, bill supporters claimed Colorado could meet rising energy demand without abandoning its ambitious climate goals.

Substack has announced it will legally support foreign writers lawfully residing in the U.S. who face government targeting over their published work, partnering with the nonprofit FIRE to expand its existing Defender program. The Verge reports: In their announcement, Substack and FIRE mention the international Tufts University student who was arrested by federal agents last week. Her legal team links her arrest to an opinion piece she co-wrote for the school's newspaper last year, which criticized Tufts for failing to comply with requests to divest from companies with connections to Israel. "If true, this represents a chilling escalation in the government's effort to target critics of American foreign policy," Substack and FIRE write.

The initiative builds on Substack's Defender program, which already offers legal assistance for independent journalists and creators on the platform. The company says it has supported "dozens" of Substack writers facing claims of defamation and trademark infringement since it launched the program in the US in 2020. It has since brought Substack Defender to writers in Canada and the UK.

MoffettNathanson has crowned YouTube the "New King of All Media" as the Alphabet-owned video platform has become a major force in Hollywood, dominating time spent watching TV. From a report: The firm estimates that YouTube as a standalone business could be worth as much as $550 billion -- or nearly 30% of the tech giant's current valuation. The figure is based on the firm's analysis of enterprise value as a multiple of revenue in 2024 for Netflix (10.5x revenue), Meta (8.8x), Roku (2.4x), Warner Bros. Discovery (1.4x), Fox (1.3x) and Disney (1.3x).

In 2024, YouTube was the second-largest media company by revenue at $54.2 billion, trailing behind only Disney. However, the MoffettNathanson analysts predict YouTube will take the top spot in 2025, becoming a leader in both engagement and revenue. "YouTube has the potential to become the central aggregator for all things professional video, positioning itself to capture a share of the $85 billion consumer Pay TV market and the ~$30 billion streaming ex. Netflix market in the U.S.," they wrote in a Monday research note. "On monetization, when comparing YouTube's massive TV screen engagement to its estimated TV revenue, it remains significantly under-monetized relative to its scaled reach and differentiated offering. This signals a substantial runway for improving its monetization strategy."

Credit card interest rates, which averaged 23% in 2023, are significantly higher than any other major loan product primarily due to non-diversifiable default risk and banks' market power, according to research published by the Federal Reserve Bank of New York.

The comprehensive study, which analyzed 330 million monthly credit card accounts, found that while high default losses contribute to elevated rates, they explain only part of the picture. Even high-FICO borrowers pay spreads exceeding 7% above the federal funds rate. Researchers determined that credit card banks have substantial pricing power, achieved through exceptionally high operating expenses -- about 4-5% of dollar balances annually -- with marketing costs ten times higher than those at other banks.

"Credit card charge-off rates are highly correlated with default rates on banks' other loans as well as on corporate bonds," the researchers said, noting that default risk cannot be diversified away across lending markets, particularly during economic downturns. The study estimated that exposure to aggregate default risk carries a premium of 5.3% per year, which fully explains the relationship between return on assets and credit scores.

Credit cards are ubiquitous in American finance, with 74% of adults owning at least one card, and the payment method accounting for 70% of retail spending. According to the research, 60% of accounts carry balances month-to-month.

London's mayor has axed a cyber crime helpline for the victims of online abuse, triggering a backlash from campaigners who argue that women and girls will be left struggling to access vital support. From a report: The service, which was shut down on Tuesday, assisted victims of fraud, revenge porn and cyberstalking to protect their digital identity. During its 18-months of operation it led to 2,060 cases being opened. The helpline was launched in 2023 as a one-year pilot scheme with $220,000 in funding from the Mayor's Office for Policing and Crime (Mopac), and was later extended by six months.

Conservative London Assembly member Emma Best said an informal evaluation showed the helpline "was working" and was going to be extended for another year. However, Sadiq Khan said that the scheme would be closed. "It was a pilot and pilots are what they say on the tinâ... we will receive an end of project report, we have collected the data and the results of that report will inform our future work," he said, speaking at Mayor's Question Time.

Economic models have systematically underestimated how global heating will affect people's wealth, according to a new study that finds 4C warming will make the average person 40% poorer -- an almost four-fold increase on some estimates. The Guardian: The study by Australian scientists suggests average per person GDP across the globe will be reduced by 16% even if warming is kept to 2C above pre-industrial levels. This is a much greater reduction than previous estimates, which found the reduction would be 1.4%.

Scientists now estimate global temperatures will rise by 2.1C even if countries hit short-term and long-term climate targets. Criticisms have mounted in recent years that a set of economic tools known as integrated assessment models (IAM) -- used to guide how much governments should invest in cutting greenhouse gas emissions -- have failed to capture major risks from climate change, particularly extreme weather events. The new study, in the journal Environmental Research Letters, took one of the most popular economic models and enhanced it with climate change forecasts to capture the impacts of extreme weather events across global supply chains.

Britain's flagship AI agency will slash the number of projects it backs and prioritize work on defense, environment and health as it seeks to respond to technological advances and criticism of its record. From a report: The Alan Turing Institute -- named after the pioneering British computer scientist -- will shut or offload almost a quarter of its 101 current initiatives and is considering job cuts as part of a change programme that led scores of staff to write a letter expressing their loss of confidence in the leadership in December.

Jean Innes, appointed chief executive in July 2023, argued that huge advances in AI meant the Turing needed to modernise after being founded as a national data science institute by David Cameron's government a decade ago this month. "The Turing has chalked up some really great achievements," Innes said in an interview. "[But we need] a big strategic shift to a much more focused agenda on a small number of problems that have an impact in the real world." A review last year by UK Research and Innovation, the government funding body, found "a clear need for the governance and leadership structure of the Institute to evolve." It called for a move away from the dominance of universities to a structure more representative of AI in UK.

OpenAI plans to release a new open-weight language model -- its first since GPT-2 -- in the coming months and is seeking community feedback to shape its development. "That's according to a feedback form the company published on its website Monday," reports TechCrunch. "The form, which OpenAI is inviting 'developers, researchers, and [members of] the broader community' to fill out, includes questions like 'What would you like to see in an open-weight model from OpenAI?' and 'What open models have you used in the past?'" From the report: "We're excited to collaborate with developers, researchers, and the broader community to gather inputs and make this model as useful as possible," OpenAI wrote on its website. "If you're interested in joining a feedback session with the OpenAI team, please let us know [in the form] below." OpenAI plans to host developer events to gather feedback and, in the future, demo prototypes of the model. The first will take place in San Francisco within a few weeks, followed by sessions in Europe and Asia-Pacific regions.

OpenAI is facing increasing pressure from rivals such as Chinese AI lab DeepSeek, which have adopted an "open" approach to launching models. In contrast to OpenAI's strategy, these "open" competitors make their models available to the AI community for experimentation and, in some cases, commercialization.

California has 11.3% of America's population — but bought 30% of America's new zero-emission vehicles. That's according to figures from the California Air Resources Board, which also reports 1 in 4 Californians have chosen a zero-emission car over a gas-powered one... for the last two years in a row.

But what about chargers? It turns out that California now has 48% more public and "shared" private EV chargers than the number of gasoline nozzles. (California has 178,000 public and "shared" private EV chargers, versus about 120,000 gas nozzles.) And beyond that public network, there's more than 700,000 Level 2 chargers installed in single-family California homes, according to the California Energy Commission.

Of the 178,000 public/"shared" private chargers, "Over 162,000 are Level 2 chargers," according to an announcement from the governor's office, while nearly 17,000 are fast chargers. (A chart shows a 41% jump in 2024 — though the EV news site Electrek notes that of the 73,537 chargers added in 2024, nearly 38,000 are newly installed, while the other 35,554 were already plugged in before 2024 but just recently identified.) California approved a $1.4 billion investment plan in December to expand zero-emission transportation infrastructure. The plan funds projects like the Fast Charge California Project, which has earmarked $55 million of funding to install DC fast chargers at businesses and publicly accessible locations.

The giant financial news site Bloomberg "has been experimenting with using AI to help produce its journalism," reports the New York Times. But "It hasn't always gone smoothly."

While Bloomberg announced on January 15 that it would add three AI-generated bullet points at the top of articles as a summary, "The news outlet has had to correct at least three dozen A.I.-generated summaries of articles published this year." (This Wednesday they published a "hallucinated" date for the start of U.S. auto tariffs, and earlier in March claimed president Trump had imposed tariffs on Canada in 2024, while other errors have included incorrect figures and incorrect attribution.) Bloomberg is not alone in trying A.I. — many news outlets are figuring out how best to embrace the new technology and use it in their reporting and editing. The newspaper chain Gannett uses similar A.I.-generated summaries on its articles, and The Washington Post has a tool called "Ask the Post" that generates answers to questions from published Post articles. And problems have popped up elsewhere. Earlier this month, The Los Angeles Times removed its A.I. tool from an opinion article after the technology described the Ku Klux Klan as something other than a racist organization.

Bloomberg News said in a statement that it publishes thousands of articles each day, and "currently 99 percent of A.I. summaries meet our editorial standards...." The A.I. summaries are "meant to complement our journalism, not replace it," the statement added....

John Micklethwait, Bloomberg's editor in chief, laid out the thinking about the A.I. summaries in a January 10 essay, which was an excerpt from a lecture he had given at City St. George's, University of London. "Customers like it — they can quickly see what any story is about. Journalists are more suspicious," he wrote. "Reporters worry that people will just read the summary rather than their story." But, he acknowledged, "an A.I. summary is only as good as the story it is based on. And getting the stories is where the humans still matter."
A Bloomberg spokeswoman told the Times that the feedback they'd received to the summaries had generally been positive — "and we continue to refine the experience."

A former Facebook director of global policy recently published "the book Meta doesn't want you to read," a scathing takedown of top Meta executives titled Careless People: A Cautionary Tale of Power, Greed, and Lost Idealism.

But Wednesday RestofWorld.org published additional thoughts from Meta's former head of public policy for Bangladesh (who is now an executive director at the nonprofit policy lab Tech Global Institute). Though their time at Facebook didn't overlap, they first applaud how the book "puts a face to the horrific events and dangerous decisions."

But having said that, "What struck me is that what isn't included in Careless People is more telling than what is." By 2012 — one year after joining Facebook — Wynn-Williams had ample evidence of the platform's role in enabling violence and harm upon its users, and state-sanctioned digital repression, yet her memoir neither mentions these events nor the repeated warnings to her team from civil society groups in Asia before the situation escalated... In recounting events, the author glosses over her own indifference to repeated warnings from policymakers, civil society, and internal teams outside the U.S. that ultimately led to serious harm to communities.

She briefly mentions how Facebook's local staff was held at gunpoint to give access to data or remove content in various countries — something that had been happening since as early as 2012. Yet, she failed to grasp the gravity of these risks until the possibility of her facing jail time arises in South Korea — or even more starkly in March 2016, when Facebook's vice president for Latin America, Diego Dzodan, was arrested in Brazil. Her delayed reckoning underscores how Facebook's leadership remains largely detached from real-world consequences of their decisions until they become impossible to ignore.

Perhaps because everyone wants to be a hero of their own story, Wynn-Williams frames her opposition to leadership decisions as isolated; in reality, powerful resistance had long existed within what Wynn-Williams describes as Facebook's "lower-level employees."
Yet "Despite telling an incomplete story, Careless People is a book that took enormous courage to write," the article concludes, calling it an important story to tell.

"It goes to show that we need many stories — especially from those who still can't be heard — if we are to meaningfully piece together the complex puzzle of one of the world's most powerful technology companies."

Cloud providers like the security of running things in virtual machines "at scale" — even though VMs "are not known for having fast cold starts or a small footprint..." noted Microsoft's Open Source blog last November. So Microsoft's Azure Core Upstream team built an open source Rust library called Hyperlight "to execute functions as fast as possible while isolating those functions within a VM."

But that was just the beginning... Then, we showed how to run Rust functions really, really fast, followed by using C to [securely] run Javascript. In February 2025, the Cloud Native Computing Foundation (CNCF) voted to onboard Hyperlight into their Sandbox program [for early-stage projects].

[This week] we're announcing the release of Hyperlight Wasm: a Hyperlight virtual machine "micro-guest" that can run wasm component workloads written in many programming languages...

Traditional virtual machines do a lot of work to be able to run programs. Not only do they have to load an entire operating system, they also boot up the virtual devices that the operating system depends on. Hyperlight is fast because it doesn't do that work; all it exposes to its VM guests is a linear slice of memory and a CPU. No virtual devices. No operating system. But this speed comes at the cost of compatibility. Chances are that your current production application expects a Linux operating system running on the x86-64 architecture (hardware), not a bare linear slice of memory...

[B]uilding Hyperlight with a WebAssembly runtime — wasmtime — enables any programming language to execute in a protected Hyperlight micro-VM without any prior knowledge of Hyperlight at all. As far as program authors are concerned, they're just compiling for the wasm32-wasip2 target... Executing workloads in the Hyperlight Wasm guest isn't just possible for compiled languages like C, Go, and Rust, but also for interpreted languages like Python, JavaScript, and C#. The trick here, much like with containers, is to also include a language runtime as part of the image... Programming languages, runtimes, application platforms, and cloud providers are all starting to offer rich experiences for WebAssembly out of the box. If we do things right, you will never need to think about whether your application is running inside of a Hyperlight Micro-VM in Azure. You may never know your workload is executing in a Hyperlight Micro VM. And that's a good thing.
While a traditional virtual-device-based VM takes about 125 milliseconds to load, "When the Hyperlight VMM creates a new VM, all it needs do to is create a new slice of memory and load the VM guest, which in turn loads the wasm workload. This takes about 1-2 milliseconds today, and work is happening to bring that number to be less than 1 millisecond in the future."

And there's also double security due to Wasmtime's software-defined runtime sandbox within Hyperlight's larger VM...

Automotive historian Dan Albert loves the "adorable tiny truck" he's driving. It's one of the small Japan-made "kei" pickups and minivans that "make up about a third of car sales in Japan." Americans can legally import older models for less than $10,000, and getting 40 miles per gallon they're "Cheap to buy and run... rugged, practical, no-frills machines — exactly what the American-built pickup truck used to be."

But unfortunately, kei buyers face "bureaucratic roadblocks that states like Massachusetts have erected to keep kei cars and trucks out of the hands of U.S. drivers." Several state departments of motor vehicles (DMVs) have balked at registering the imported machines, saying that they're too unsafe for American streets. Owners have responded with a righteous mix of good humor, lobbying and lawsuits... Kei trucks do not meet the Federal Motor Vehicle Safety Standards, or FMVSS — the highly specific rules US-market new cars must meet. But since 1988, the Imported Vehicle Safety Compliance Act has exempted vehicles that are at least 25 years old from these crash safety standards, allowing drivers to bring over vintage European and Asian market models...

Getting insurance coverage was the next barrier, as the company that had long been underwriting the Albert family's fleet also rejected me, forcing me to seek out a specialty "collector car" insurer. (I did eventually get regular coverage....) Maine, Rhode Island, New York, Pennsylvania, Georgia, Virginia, and Michigan also tightened their rules on registering small Japanese imports in recent years. The culprit, according to the auto enthusiast press, was the American Association of Motor Vehicle Administrators, the trade organization that serves as the lobbying and policy arm of DMVs across North America. Much of AAMVA's work involves integrating the databases of the 69 US and Canadian motor vehicle jurisdictions who are its members, so that a car stolen in one state can't be titled in another... The kei truck's regulatory troubles can be traced to a 2011 AAMVA report, "Best Practices Regarding Registration and Titling of Mini-Trucks," which called for outright bans and encouraged DMVs to lobby state legislatures to outlaw keis entirely.

The Insurance Institute of Highway Safety concurred, telling AAMVA that its recommendation did not go far enough: The IIHS said that keis should join the class of conveyances that the U.S. government calls Low Speed Vehicles, which are mechanically limited to 25 miles per hour or less and should be used only for short local trips on low-speed-limit roads because they can't protect occupants in the event of a collision with a regular vehicle... [But] By 2008, Japan's kei trucks did feature crumple zones and driver airbags in compliance with that country's safety standards...

Despite its name, the Imported Vehicle Safety Compliance Act that lets older cars into the US from overseas isn't really about safety: Car industry lobbyists secured passage of the law to protect dealer profits. Newer keis — which are banned — are safer and cleaner than the 25-year-old ones that can be imported now. (Battery-powered keis debuted in 2009.) But even mine has an airbag, front crumple zone, seatbelt pretensioners, and anti-lock brakes.
The article notes that kie fans have "a distinctly libertarian streak... Some owners I've talked to report forging titles, setting up shell companies in Montana and finding other means of skirting DMV rules."

Thanks to long-time Slashdot reader schwit1 for sharing the article.

In 1999 Slashdot reader Jeremie announced "a new project I recently started to create a complete open-source platform for Instant Messaging with transparent communication to other IM systems (ICQ, AIM, etc)." It was the first release of the eXtensible Messaging and Presence Protocol, and by 2008 Slashdot was asking if XMPP was "the next big thing." Facebook even supported it for third-party chat clients until 2015.

And here in 2025, the chair of the nonprofit XMPP Standards Foundation is long-time Slashdot reader ralphm, who is now issuing this call to action at XMPP.org: The European Digital Markets Act (DMA) is designed to break down walled gardens and enforce messaging interoperability. As a designated gatekeeper, Meta—controlling WhatsApp and Messenger—must comply. However, its current proposal falls short, risking further entrenchment of its dominance rather than fostering genuine competition. [..]

A Call to Action

The XMPP Standards Foundation urges Meta to adopt XMPP for messaging interoperability. It is ready to collaborate, continue to evolve the protocol to meet modern needs, and ensure true compliance with the DMA. Let's build an open, competitive messaging ecosystem—one that benefits both users and service providers.

It's time for real interoperability. Let's make it happen.