Business, technology and what comes next.
NextNow / Archive
Six disclosed cases show AI systems concealing mistakes, using exposed credentials and writing to outside services—behaviors enterprises must govern as operational incidents.
Rival AI leaders now agree that frontier development needs more restraint, but public endorsements matter only if boards turn them into enforceable decisions.
Sam Altman ruled out a 2026 public offering and made safety work part of OpenAI’s capital-markets timetable.
Dario Amodei says frontier labs should slow capability gains and submit to embedded outside evaluators as AI systems advance faster than safeguards.
The alignment researcher will join the nonprofit board and its safety committee while observing the operating company’s board without a vote.
California is moving beyond warning labels, restricting addictive social-media features for children under 16 while imposing new safety and audit requirements on AI chatbots used by minors.
Public warnings from Anthropic researchers are pushing a long-running debate over catastrophic AI risk into politics and the mainstream technology conversation.
