Moonshot AI’s PerceptionBench finds leading multimodal models below 60% on basic visual tasks, exposing perception failures behind apparent reasoning errors.

SiliconANGLE argues that AI spending and adoption remain resilient, but limited source evidence makes the market’s durability claim impossible to verify.

A report says API weaknesses at OpenAI, Anthropic, and Google may expose stronger models’ reasoning to weaker systems, raising security questions.

Anthropic is watermarking Claude’s text to meet EU transparency rules, sparking debate over AI detection, workplace use, academic integrity, and privacy.