AI companies are now asking governments before releasing powerful new tools

Anthropic delayed releasing Claude Mythos, a tool designed to find security flaws, because European regulators raised concerns. OpenAI, Google DeepMind, and Anthropic are all coordinating release dates with governments now, instead of launching and facing rules later. This marks a shift in how power works: companies consult regulators before releasing, not after. It is unclear whether this keeps everyone safer or just gives governments more control.