AI
AI That Predicts You Still Needs to Show It Can Adapt
A USC study and NIST guidance point to a sharper test for AI personalization: whether new evidence changes what an assistant actually does.
Section
AI
Artificial Intelligence (AI)
From the newsroom
Stories
AI
XPENG’s IRON Production Line Raises the Bar for Robot Evidence
XPENG has moved its IRON humanoid robot onto a production line. The more consequential question is what evidence…
AI
GPT-6 Astra Makes the Harness Part of the Performance Story
OpenAI’s GPT-6 Astra release shows why an AI agent’s memory, tools, permissions, and oversight now matter as much…
AI
Gemini’s Six-Week Release Sprint Turns Model Choice Into Change Management
Three Gemini Flash generations arrived in 43 days, turning model selection into a continuous discipline of evaluation, staged…
AI
AI Labs Are Turning Cyber Evaluations Into Layered Security Systems
Anthropic and OpenAI are converging on layered isolation, monitoring and stop conditions after agentic cyber tests crossed intended…
AI
AI Literacy Becomes Part of Every Major at ASU
Arizona State University is putting discipline-specific AI literacy across every college, shifting the test from tool access to…
AI
Wildfire Detection AI Is Fastest When Humans Stay in the Loop
California’s AI wildfire cameras create an earlier warning opportunity, but coverage, human confirmation, and delivery still determine operational…
AI
AI Safety Research Can Now Be Automated—Within Narrow Bounds
A new study shows AI agents can improve models on defined safety benchmarks, while exposing oversight as the…
AI
Double-Blind AI Testing Hides the Exam From Both Sides
A new confidential-computing pilot separates model owners from benchmark owners, strengthening the chain of custody behind closed AI…