Tech

A new benchmark called NRT-Bench tests what happens when LLM agents run a simulated nuclear power plant and adversaries try to trick them into triggering a meltdown. The results, published on arXiv on June 18, show that adaptive multi-turn attacks reliably push AI operators past safety limits (arXiv:2606.20408, June 18). The Setup The researchers built […]

A new benchmark called TRAP has exposed a fundamental privacy vulnerability in AI agents: every model tested leaks private information when prompted adversarially, and no amount of prompt engineering can fully fix it. The results, published on arXiv on June 17, point to a structural problem that may require architectural changes, not better instructions (arXiv:2606.18996, […]

A data breach at a Texas state government department allowed hackers to steal the driver’s license information and passport numbers of more than 3 million people, according to the state’s attorney general (TechCrunch, June 18). The incident is one of the largest data breaches to affect the state this year. ### What Happened The Texas […]

OpenAI published a new pre-release safety method on June 16 called Deployment Simulation. The idea is deceptively simple: before a model ships, replay real user conversations through it and measure how it behaves. The approach has already influenced GPT-5-series thinking launches and caught novel misalignment patterns that traditional evaluations missed (MarkTechPost, June 16; OpenAI paper, […]

Bernie Sanders has introduced legislation that would take roughly $7 trillion from America’s largest AI companies and hand the public a 50% ownership stake. The American AI Sovereign Wealth Fund Act, unveiled June 18, goes further than any previous proposal in giving citizens voting power over corporate decisions, not just a share of profits (Ars […]

It has been nearly a week since the Trump administration ordered Anthropic to pull its most advanced models offline. Claude Mythos 5 and Fable 5 remain suspended. Anthropic leaders have flown to Washington, met with White House officials, and negotiated for days. And yet no one can say exactly what the company did wrong (Wired, […]

Microsoft has quietly built a booming business selling OpenAI’s GPT models to Chinese technology giants, even as OpenAI and Anthropic refuse to sell their AI directly into the country over fears of intellectual property theft and misuse. The arrangement, detailed in reporting by Bloomberg published June 18, hands Microsoft a position no other American AI […]

Intel Foundry announced that its 18A-P process, the first performance enhancement to the Intel 18A node family, has entered risk production. The milestone, disclosed at the 2026 VLSI Symposium alongside six research papers, marks the company’s most detailed public accounting of how its gate-all-around (GAA) transistors and backside power delivery architecture perform in real designs. […]

When a chip fabrication plant discovers a pattern of low-yielding wafers, finding the root cause can take weeks. Engineers set up expensive Design of Experiment tests, dedicate operators to track wafers manually, and wait for results. And the test lot may not even reproduce the problem. A technique called wafer randomization promises to change that, […]

Days before the Trump administration forced Anthropic to take its most powerful AI models offline globally, the White House ordered the company to revoke a single customer’s access: SK Telecom, South Korea’s largest telecommunications provider. The reason, according to administration officials, was the company’s alleged ties to China. The SK Telecom story is the missing […]