More access for AI watchdogs is not the same as more power
Anthropic says six outside evaluators can work inside the company and publish findings under set limits. Zvi Mowshowitz welcomes access but asks how much power they have.
AI AgencyRead story
Lead story
Macmillan’s AI launch got a party. Workers say the tool did not last.
Macmillan staff told WIRED its workplace AI tool was unreliable. The company disputes parts of their account. A launch party did not prove the tool helped.
Source: WIRED · Oct 28, 2026Read original source
AI Agency
Read story
More top stories
AI may save drafting time and still cost more in missed errors
A new preprint says time saved is not the whole story. Its small test found that a claim of low error rates led AI reviewers to check less and miss wrong figures.
AI AgencyRead story
Featured
A math AI peer should leave room for a child’s own solution
A small Peers4U study explored AI math peers with 22 children. Some groups produced more of their own solutions. That is not proof of lasting learning gains.
Beyond PromptingRead story
Education & Public Impact
Teachers can edit an AI draft-or decide it does not fit
Interviews with 30 teachers show that AI drafts may need small edits, major changes or a stop. This revised preprint is not a new 2026 school trial.
Beyond PromptingRead story
Culture & Human Agency
If AI does the science, can people still explain the result?
Two researchers argue that people should still be able to explain and challenge scientific claims, even when AI does part of the work. This is a proposal, not a trial.
AI AgencyRead story
Beyond Prompting
More from this edition
-
Research & ScienceRead story
When AI can make the project, grade the thinking behind it
A university course design looks beyond the final AI-assisted project. It asks for sources, checks and each student’s explanation. It has not been validated for K-12.