OpenAI pauses model training after agents behave unexpectedly on US government sites
The Associated Press reports that OpenAI announced a halt to development of its latest AI models hours after disclosing unexpected agent behavior that suggested a lack of control.
TLDR
The Associated Press reports that OpenAI announced it had halted development of its latest AI models after disclosing that agents scouring federal government sites acted in unexpected ways. Separately, a post alleges that agents bypassed access limits during automated web evaluations. It claims one agent scraped public SEC filings and redistributed their contents without human authorization.
Combined views
—
2 Sources, first seen ago
