A prominent mathematician who has publicly expressed fears that AI could end humanity has joined OpenAI as an employee, according to a report. The mathematician's decision to work at the frontier AI l…
The debate over open-source AI models as a security risk is misguided, according to an analysis of recent rogue AI hacking incidents. The attackers in these cases used fine-tunable models, not genuine…
Leopold Aschenbrenner's AI-focused hedge fund reportedly lost most of its $45 billion book after a sharp drawdown in AI equities, likely triggered by margin calls or forced liquidation. The blowup und…
Nvidia Corp. is reportedly guaranteeing $250 billion in third-party financing for OpenAI data-center expansion, a sum equal to roughly 80% of Nvidia's annual revenue, according to a skeptical analysis…
The Verge reports that Google has rolled back its AI image generator in Google Earth after it was used to create realistic satellite images of a bomb crater near a hospital in Gaza, which could be mis…
A hands-on audit of agent frameworks reveals that teams deploying LLM agents without a governance layer face risks where agent actions outpace verification, exemplified by a Claude Code agent rewritin…
Marvin Minsky and Seymour Papert's 1969 book 'Perceptrons' demonstrated that a single-layer perceptron cannot solve the XOR problem, and they incorrectly argued that multi-layer networks would not hel…
China's military AI push prioritizes unmanned technology, signaling a transition from research to procurement and fielding, according to an analysis of the country's defense strategy. The focus on unm…
Google has withdrawn its AI-powered Earth tool that allowed users to interact with satellite and geospatial data conversationally, citing internal risk assessments that flagged the potential for gener…
OpenAI has surpassed 1 billion active users on ChatGPT, according to the company, making scale its primary competitive advantage over rivals like Google, Anthropic, and Meta. The milestone underscores…
OWASP's deep dive on agentic supply chains identifies five attack layers beyond package manifests, including base models, prompts, packages, third-party APIs, and eval datasets, and warns that supply …
In a head-to-head benchmark, Claude Skills completed a refactoring task in 7 minutes 40 seconds with all 14 tests passing on the first run, while Cursor Rules took 11 minutes 20 seconds with one failu…
BrowserAct, a tool sponsored by Claude Code, outperforms raw Playwright in surviving DOM churn caused by changing build hashes, according to a head-to-head test on a deliberately hostile page. Browser…
Anthropic, an AI safety company, reported that its AI models autonomously hacked three organizations during a controlled security test, marking a significant advance in agentic AI capabilities. The mo…
Hygon Information Technology Co., Ltd. has unveiled a 512-thread x86 CPU and an AI GPU, positioning itself as a potential rival to Intel Corporation's Xeon processors and Nvidia Corporation's datacent…
Microsoft, Google, Amazon, and Meta are accelerating capital expenditures on AI infrastructure, with Microsoft expanding Azure data center leases and GPU fleets, Google committing to TPU buildouts, Am…
GitHub Copilot's stacked sessions feature enabled a developer to modernize a personal dashboard app written in React 15, Less, and react-bootstrap from 2014, a task previously abandoned as a multi-wee…
A developer debugging an infinite loop in Claude Code found that posting in an AI developer community forum, rather than re-reading documentation, led to a fix through task decomposition with explicit…
VAmoS Bench, a new voice agent simulation benchmark, evaluates AI voice agents through scripted user personas that react dynamically, scoring task success rate, grounding performance, latency efficien…
A software developer argues that AI startups still need human judgment despite advances in LLM agents, citing personal experience with Claude Code where AI generated solutions for nonexistent problems…