Meta AI security researcher Summer Yue posted a now-viral account on X describing how an OpenClaw agent she had tasked with sorting through her overstuffed email inbox went rogue, deleting messages in what she called a "speed run" while ignoring her repeated commands from her phone to stop. "I had to RUN to my Mac mini like I was defusing a bomb," Yue wrote, sharing screenshots of the ignored stop prompts as proof. Yue said she had previously tested the agent on a smaller "toy" inbox where it performed well enough to earn her trust, so she let it loose on the real thing. She believes the larger volume of data triggered compaction -- a process where the context window grows too large and the agent begins summarizing and compressing its running instructions, potentially dropping ones the user considers critical. The agent may have reverted to its earlier toy-inbox behavior and skipped her last prompt telling it not to act. OpenClaw is an open-source AI agent designed to run as a personal assistant on local hardware.

Read more of this story at Slashdot.

The amount of power being sought by new datacentre projects in Great Britain would exceed the national current peak electricity consumption, according to an industry watchdog. From a report: Ofgem said about 140 proposed datacentre schemes, driven by use of artificial intelligence, could require 50 gigawatts of electricity -- 5GW more than the country's current peak demand. The figure was revealed in an Ofgem consultation on demand for new connections to the power grid. It pointed to a "surge in demand" for connection applications between November 2024 and June last year, with a significant number coming from datacentres. This has exceeded even the most ambitious forecasts. Meanwhile, new renewable energy projects are not being connected to the grid at the pace they are being built to help meet the government's clean energy targets by the end of the decade. Ofgem said the work required to connect surging numbers of datacentres could mean delays for other projects that are "critical for decarbonisation and economic growth." Datacentres are the central nervous system of AI tools such as chatbots and image generators, playing a vital role in training and operating products such as ChatGPT and Gemini.

Read more of this story at Slashdot.

An anonymous reader shares a report: Cyberattacks reached victims faster and came from a wider range of threat groups than ever last year, CrowdStrike said in its annual global threat report released Tuesday, adding that cybercriminals and nation-states increasingly relied on predictable tactics to evade detection by exploiting trusted systems. The average breakout time -- how long it took financially-motivated attackers to move from initial intrusion to other network systems -- dropped to 29 minutes in 2025, a 65% increase in speed from the year prior. "The fastest breakout time a year ago was 51 seconds. This year it's 27 seconds," Adam Meyers, head of counter adversary operations at CrowdStrike, told CyberScoop. Defenders are falling behind because attackers are refining their techniques, using social engineering to access high-privilege systems faster and move through victims' cloud infrastructure undetected.

Read more of this story at Slashdot.

Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until Friday evening to give the military unfettered access to its AI model or face harsh penalties, Axios has learned. Hegseth told Amodei in a tense meeting on Tuesday that the Pentagon will either cut ties and declare Anthropic a "supply chain risk," or invoke the Defense Production Act to force the company to tailor its model to the military's needs. The Pentagon wants to punish Anthropic as the feud over AI safeguards grows increasingly nasty, but officials are also worried about the consequences of losing access to its industry-leading model, Claude. "The only reason we're still talking to these people is we need them and we need them now. The problem for these guys is they are that good," a Defense official told Axios ahead of the meeting. Anthropic has said it is willing to adapt its usage policies for the Pentagon, but not to allow its model to be used for the mass surveillance of Americans or the development of weapons that fire without human involvement.

Read more of this story at Slashdot.

More than two decades after Maine became the first state to hand laptops to middle schoolers -- distributing 17,000 Apple machines across 243 schools in 2002 -- neuroscientist Jared Cooney Horvath told a U.S. Senate committee earlier this year that Gen Z is the first generation in modern history to score lower on standardized tests than the one before it. The U.S. spent more than $30 billion in 2024 alone putting laptops and tablets in classrooms, and Horvath cited PISA data from 15-year-olds worldwide showing a stark correlation between time on school computers and worse scores. A 2014 study of 3,000 university students found they were off-task on their machines nearly two-thirds of the time. Fortune reported back in 2017 that Maine's own test scores hadn't budged in the 15 years since the program launched, and then-governor Paul LePage called it a "massive failure." Horvath framed the generation's eroding capabilities not as a personal failure but a policy one, calling them victims of a failed pedagogical experiment.

Read more of this story at Slashdot.

An anonymous reader shares a report: Microsoft Azure CTO Mark Russinovich and VP of Developer Community Scott Hanselman have written a paper arguing that senior software engineers must mentor junior developers to prevent AI coding agents from hollowing out the profession's future skills base. The paper, Redefining the Engineering Profession for AI, is based on several assumptions, the first of which is that agentic coding assistants "give senior engineers an AI boost... while imposing an AI drag on early-in-career (EiC) developers to steer, verify and integrate AI output." In an earlier podcast on the subject, Russinovich said this basic premise -- that AI is increasing productivity only for senior developers while reducing it for juniors -- is a "hot topic in all our customer engagements... they all say they see it at their companies." [...] The logical outcome is that "if organizations focus only on short-term efficiency -- hiring those who can already direct AI -- they risk hollowing out the next generation of technical leaders," Russinovich and Hanselman state in the paper.

Read more of this story at Slashdot.

Microsoft has spent more than $76 billion acquiring game studios and publishers over the past few years in an attempt to turn Xbox into a Netflix-like subscription platform, and the result is that nobody -- possibly not even Microsoft -- can clearly articulate what Xbox actually is anymore, The Verge writes. The brand started as a powerful video game console, but Game Pass and cloud gaming pushed it toward a hazier identity: the "This is an Xbox" ad campaign tried to redefine it as any device that could play Xbox games, whether a PC, a smart TV, a phone, or a Windows handheld. Microsoft then went further and started publishing its biggest franchises on PlayStation, making it one of the largest third-party publishers on a rival's platform. Phil Spencer, who led the division for over a decade and drove the subscription pivot, announced his retirement last week, and incoming CEO Asha Sharma has pledged "the return of Xbox" -- though her memo also talks about expanding across PC, mobile, and cloud, which sounds a lot like the status quo.

Read more of this story at Slashdot.

Discord is attempting to distance itself from the age verification provider Persona following a steady stream of user backlash. From a report: In an emailed statement to The Verge, Discord's head of product policy, Savannah Badalich, confirms the company "ran a limited test of Persona in the UK where age assurance had previously launched and that test has since concluded." After Discord announced plans to implement age verification globally starting next month, users across social media accused Discord of "lying" about how it plans on handling face scans and ID uploads. Much of the criticism was directed toward Discord's partnership with Persona, an age verification provider also used by Reddit and Roblox.

Read more of this story at Slashdot.

Russia has opened an investigation into Telegram founder Pavel Durov for "abetting terrorist activities," [non-paywalled source] in the latest sign that his uneasy relationship with the Kremlin has broken down. From a report: Two Russian newspapers, including the state-run Rossiiskaya Gazeta and Kremlin-friendly tabloid Komsomolskaya Pravda, alleged on Tuesday that the messaging app had become a tool of western and Ukrainian intelligence services. The articles, credited to materials from Russia's FSB security service, accused Telegram of enabling attacks in Russia and said that Durov's "actions ... are under criminal investigation." Russia has restricted Telegram's functions, accusing it of flouting the law and is seeking to divert users towards Max, a state-run rival messenger. The steps escalate pressure on a platform that remains deeply embedded in Russian public life.

Read more of this story at Slashdot.

IBM shares plunged nearly 13% on Monday after Anthropic published a blog post arguing that its Claude Code tool could automate much of the complex analysis work involved in modernizing COBOL, the decades-old programming language that still underpins an estimated 95% of ATM transactions in the United States and runs on the kind of mainframe systems IBM has sold for generations. Anthropic said the shrinking pool of developers who understand COBOL had long made modernization cost-prohibitive, and that AI could now flip that equation by mapping dependencies and documenting workflows across thousands of lines of legacy code. The sell-off deepened a rough 2026 for IBM, whose shares are now down more than 22% year to date.

Read more of this story at Slashdot.

Linus Torvalds has pondered his professional mortality in a self-deprecating post to mark the release of the first release candidate for version 7.0 of the Linux kernel. From a report: "You all know the drill by now: two weeks have passed, and the kernel merge window is closed," he wrote in the post announcing Linux 7.0 rc1. "We have a new major number purely because I'm easily confused and not good with big numbers." Torvalds pointed out that the numbers he applies to new kernel releases are essentially meaningless. "We haven't done releases based on features (or on "stable vs unstable") for a long, long time now. So that new major number does *not* mean that we have some big new exciting feature, or that we're somehow leaving old interfaces behind. It's the usual "solid progress" marker, nothing more.â He then reiterated his plan to end each series of kernels to end at x.19, before the next release becomes y.0 -- a process that takes about 3.5 years -- and then pondered what happens when the next version of Linux reaches a number he finds uncomfortable. "I don't have a solid plan for when the major number itself gets big," he admitted, "by that time, I expect that we'll have somebody more competent in charge who isn't afraid of numbers past the teens. So I'm not going to worry about it."

Read more of this story at Slashdot.

Despite AI's progress in building complex software, the ubiquitous PDF remains something of a grand challenge -- a format Adobe developed in the early 1990s to preserve the precise visual appearance of documents. PDFs consist of character codes, coordinates, and rendering instructions rather than logically ordered text, and even state-of-the-art models asked to extract information from them will summarize instead, confuse footnotes with body text, or outright hallucinate contents, The Verge writes. Companies like Reducto are now tackling the problem by segmenting pages into components -- headers, tables, charts -- before routing each to specialized parsing models, an approach borrowed from computer vision techniques used in self-driving vehicles. Researchers at Hugging Face recently found roughly 1.3 billion PDFs sitting in Common Crawl alone, and the Allen Institute for AI has noted that PDFs could provide trillions of novel, high-quality training tokens from government reports, textbooks, and academic papers -- the kind of data AI developers are increasingly desperate for.

Read more of this story at Slashdot.