Artificial Intelligence

AP explains the federal pre-release vetting framework for advanced AI systems, the voluntary role of major labs, and the policy questions it raises.

WIRED examines how Chinese open models are improving fast enough to pressure U.S. labs and policymakers on access, safety, and strategy.

A coalition of more than 20 companies says open-weight models help competition and innovation, as Washington weighs whether to curb access to Chinese systems.

The answer is less about intelligence than anatomy: chimpanzees can learn symbols and gestures, but their vocal hardware does not support human speech.

Anthropic says Claude Opus 5 is available now, with notable gains on software engineering, knowledge work, and agentic tasks—while keeping the same price as Opus 4.8.

Artificial Analysis’s model rankings put Claude Opus 5 in first place on its Intelligence Index, with the site comparing models across quality, speed, cost, latency, and context window.

Ars Technica reports that scientists have adapted AlphaFold to help identify and redesign parts of gene-editing proteins linked to off-target edits.

A research paper describes using audio IoT sensors, modulation tensorgrams, and recurrent neural networks to better estimate hive strength from noisy recordings.

WIRED reports an internal policy divide over whether the US should tighten controls on increasingly capable Chinese AI systems or keep the current approach more open.

A new open-source drawing arena asked four vision models to reproduce images and follow prompts using only colored-pencil tools.