AI 2040's hopeful path, and three more stories

This issue walks through a forecasting group's best case scenario for AI, New York's plan to use AI to weed out obsolete rules, evidence that foreign state media are amplifying opposition to data centers, and a startup's proposal to settle disputes between AI agents with an AI jury.

The AI Futures Project's new plan imagines the United States and China negotiating limits on advanced chips and frontier training, joined by other countries in a group called The Consortium, with slow and openly scrutinized development leading up to superintelligence; the writer doubts we will get AIs capable of solving alignment yet unable to rebel (Joe)
Governor Kathy Hochul has directed New York State to use AI to find outdated regulations, which the writer sees as a good match for the technology while warning that spotting questionable rules is not the same as cutting the right ones, citing last year's disruption of PEPFAR (Joe)
A report from the threat intelligence firm Alethea documents Chinese, Russian and Iranian state media referring to data centers about 700 times from January to June 2026, a small share of their overall AI coverage; the firm says the opposition began in the United States and is being exploited, not manufactured, from abroad (Alana)
GenLayer proposes an Internet Court where five blockchain participants running different AI models act as a jury for agent to agent disputes, expanding until they reach a verdict; critics note that language models share training data and failure modes, so they may not be truly independent judges (Alana)
The Forbes piece behind the Internet Court story also describes marketplaces where agents hire other agents for paid work, which the writer finds unsettling though less alarming than the extinction risk some experts warn about (Alana)