Up Close
OpenAI tries to write the rules before governments do
OpenAI published a policy paper that proposes shared international technical standards and governance mechanisms for advanced AI systems, including recursive self-improving models, and calls for U.S. leadership in shaping them.
This is likely to matter less as a research view than as an attempt to move AI oversight into the operating rules of the market before those rules are imposed piecemeal by states. Recursive self-improving models - systems meant to improve their own capabilities through repeated automated cycles - are still a proposal here, but naming them in a standards paper suggests providers now want governance machinery ready before such systems are treated as ordinary software. For Microsoft and other cloud platforms, that would likely raise the value of selling governed access, logging and control layers around models, because standards are applied through infrastructure that enterprises and regulators can actually inspect.
OpenAI also created an independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study to review mathematical results emerging from its AI systems and how those results are communicated. This appears to show a second kind of control taking shape alongside model release: not only who may use powerful systems, but how claims from them are checked before they circulate into science, education and public debate.
Google launched a Workday for Google Sheets add-on that brings Ask Workday into Sheets and Slides for licensed users.