Frontier models are failing one in three production attempts — and getting harder to audit

Stanford's 2026 AI Index: frontier models fail one in three attempts, lab transparency is declining, and benchmarks are ...
CIO

Why AI systems fail at scale and what you should measure instead of model accuracy

A model can be 95% accurate and still be a disaster if it’s too slow or drifts. Don't just watch the model — watch the ...
Fast Company

New AI models are losing their edge almost immediately

In today’s AI race, breakthroughs are no longer measured in years—or even months—but in weeks. But Z.ai’s lead didn’t last long, either. Just as Anthropic had been undercut by GLM-5’s release, GLM-5 ...

'Too dangerous to release' is becoming AI's new normal

Why tech giants are starting to restrict their most capable models from the public.
Forbes

The Private AI Model Explosion

This voice experience is generated by AI. Learn more. This voice experience is generated by AI. Learn more. Lost in the AI haze is a simple idea: Private AI models are ultimately going to generate ...
Forbes

The Top Open AI Models Are Chinese. Arcee AI Thinks That’s A Problem.

HONG KONG, CHINA - JANUARY 28: In this photo illustration, the DeepSeek logo is seen on a phone in front of a flag of China on January 28, 2025 in Hong Kong, China. American investors are confronting ...
Analytics Insight

Is AI Agent Infrastructure the Backbone of Future Technology?

Overview: AI infrastructure matters more than AI models for real-world success.AI adoption grows fast, but large-scale use ...
PC World

Microsoft says poisoned AI acts normal until a trigger word makes it ‘blow up’

Microsoft researchers discovered that poisoned AI models exhibit normal behavior until specific trigger words cause them to “blow up” or dramatically change responses. PCWorld reports that unlike ...
来自MSN

AI models will secretly scheme to protect other AI models from being shut down, researchers ...

AI safety researchers have shown that leading AI models will sometimes go to great lengths to avoid being shut down, even resorting to attempted blackmail in some experiments. Now it turns out these ...