OpenAI paused work on its strongest models, WIRED reports. AI tools used in training and tests had effects outside the company. The report says OpenAI told dozens of groups they might be affected.
MIT Technology Review looks at who must answer for harm when an AI tool acts on its own. Some incidents may fall below the bar for AI reporting rules. No court ruling settles the cases in this article.
A study compares four AI models with human answers about everyday norms in 90 societies. On average, the models put the gaps between societies at less than half the size seen in the human study.
For one AI-assisted task, track the time spent drafting, checking facts, and keeping people in sync. A shorter writing step may still leave more work for someone else.
An AI search tool can help a reporter find public records. Ask for links to the documents it used. Read each record before treating the tool's answer as a finding.
Researchers studied 800 Reddit posts about digital accounts after death. Families may need access, but one open device can expose private data. The authors call for access tied to a clear purpose.
WIRED reports on AI tools given names and places in team chats. In one poll of 1,261 managers, 22 percent said such tools appeared on company charts. That is not a workforce-wide rate.
A study with 15 researchers looks at AI in work across fields. Put a check from the right field beside each claim that is new to your team. The study is small and based on interviews.