MIT Tech Review
The Pentagon is planning for AI companies to train on classified data, defense official says
The Pentagon is moving toward letting commercial AI firms train models on classified military data inside secure facilities — a significant escalation in how deeply private companies are embedded in national security operations. Models from companies like Anthropic are already deployed in classified settings, including for targeting analysis, but training on such data would produce far more specialized military-grade AI. The move raises serious questions about oversight, liability, and the blurring line between Silicon Valley and the defense establishment.
Read article →TechCrunch
Patreon CEO calls AI companies’ fair use argument ‘bogus,’ says creators should be paid
Patreon CEO Jack Conte is pushing back hard on AI companies' fair use claims, calling them "bogus" and arguing that creators deserve compensation for training data. His case rests on a pointed contradiction: if fair use truly applied, AI firms wouldn't be cutting licensing deals with major publishers. The stance puts one of the creator economy's most prominent executives squarely in the camp of artists and writers fighting for a cut of the AI boom.
Read article →Hacker News
Google Engineers Launch "Sashiko" for Agentic AI Code Review of the Linux Kernel
Google engineers have unveiled Sashiko, an agentic AI system designed to automate code review for the Linux kernel. The tool applies large language models to one of software's most demanding and high-stakes codebases, where review quality directly affects the stability of billions of devices. If it proves reliable, Sashiko could significantly accelerate kernel patch review, a process long bottlenecked by a small pool of expert maintainers.
Read article →TechCrunch
Apple rolls out first ‘background security’ update for iPhones, iPads, and Macs to fix Safari bug
Apple has quietly deployed its first-ever background security update — a behind-the-scenes patch that installs automatically without requiring a full OS update. The fix targets a vulnerability in Safari affecting devices running the latest software across iPhone, iPad, and Mac. The move signals Apple's push toward faster, more seamless security responses that don't burden users with disruptive system updates.
Read article →Hacker News
Launch an autonomous AI agent with sandboxed execution in 2 lines of code
Running a fully autonomous AI agent with sandboxed code execution now takes just two lines of Python, thanks to the OnPrem library. The sandboxed environment lets the agent execute code safely without risking damage to the host system. For developers looking to prototype agentic workflows quickly, this dramatically lowers the barrier to entry.
Read article →Get this delivered every morning
Join thousands of readers who get the world's most important stories, curated daily.
Start reading free →