A new guide explains how everyday users can download and run large language models like those from Meta and Google directly on their own Windows, macOS, or Linux machines instead of relying on cloud-based apps like ChatGPT. Running LLMs locally offers offline access and greater privacy since no data is sent to outside servers, and it avoids subscription fees, though these free models are generally less powerful and slower than paid cloud versions. The guide notes that macOS is often preferred due to Apple Silicon's unified memory architecture, and that at least 16GB of RAM is recommended for smooth performance.
Meta reached an $18 billion settlement with a coalition of U.S. states that accused it of concealing the risks its platforms posed to young users, ending a high-profile trial in California. Under the deal Meta will pay roughly $12.7 billion outright and adopt new safeguards for minors, such as a two-hour daily limit, disabled cosmetic-surgery filters and stricter age checks, while the remaining $5.3 billion depends on rivals like TikTok and YouTube adopting similar protections.
A recent legal settlement involving Meta imposes new regulatory obligations on the company even as authorities argue it wields excessive market power. Critics note the compliance requirements demand significant resources, which larger firms can absorb more easily than smaller rivals.
Researchers in Germany tested over 120 metal fans and found that attending several days of live metal performances appeared to increase their ability to withstand pain, based on cold-pain tolerance tests. The findings suggest a link between the concert experience and heightened physical resilience among attendees.
Brave has rolled out a new privacy tool letting users generate disposable email aliases when signing up for websites, avoiding exposure of their real address. Emails sent to an alias are automatically forwarded to the user's primary inbox and then deleted from Brave's servers, with account data encrypted at rest.
Regulators have expanded a nationwide recall to cover 3.6 million grill-cleaning brushes sold over roughly 17 years, after consumers accidentally swallowed loose metal bristles. Several people required medical treatment to remove bristles lodged in their throat or digestive tract.
Researchers from Meta AI and the University of Illinois Urbana-Champaign built EvoHarness-RL, a training framework that teaches a small 8-billion-parameter model to decide when to read, update, or consolidate information it gathers from its execution environment during long, multi-step tasks. Rather than following rigid, human-written scripts for tool use, the model learns to manage its own memory and error recovery, reportedly reaching performance comparable to Claude Opus 4.5 on complex agentic workflows.
Big Tech's hyperscalers—Meta, Amazon, Alphabet and Microsoft—are increasingly relying on debt issuance, equity sales, and off-balance-sheet financing arrangements to fund surging AI infrastructure investment. Capital expenditures have exceeded 100% of operating cash flow at some of these firms, prompting Amazon and Alphabet to sell bonds globally, Alphabet to issue equity, and Meta to pursue bond sales and partnerships with alternative asset managers for data center funding.
Meta released another firmware update aimed at preventing users from disabling or covering the LED light that signals when its smart glasses are recording. The update, announced by wearables chief Alex Himel, is meant to more reliably alert bystanders that photos or video are being captured, and Meta says similar updates will keep coming.
Meta is pushing an update that stops its AI glasses from recording if the indicator LED is covered mid-capture, closing a workaround where users could start filming and then hide the light afterward. Wearables executive Alex Himel confirmed the fix on Threads, saying it ensures the light reliably signals to bystanders when they're being filmed.
Generative AI tools from companies like OpenAI, Google and Meta have made it far easier to produce convincing fake audio, video and images of public figures, fueling a rise in political deepfakes. Doppel's Bobby Ford describes the trend as driven by three factors: greater volume, faster velocity and more vectors such as video, audio and email being exploited by bad actors. The piece cites the 2024 New Hampshire robocall incident, where an AI-generated fake of Joe Biden's voice urged voters to stay home, as an early warning sign for what could happen in the 2026 midterms.
A UMass Lowell Center for Public Opinion dataset, highlighted by a Reddit user following Dolly Parton's death, shows Meta CEO Mark Zuckerberg holds just 8 percent favorability among Democrats and 13 percent among Republicans. That places him near Vladimir Putin in overall unpopularity, despite Zuckerberg's recent efforts to align himself politically with Donald Trump.
GoKawiil