MAIN CONTENT
Three directions within one race
Today’s map of opinion no longer follows familiar lines between political parties or between government and business. One camp believes the capabilities of frontier models are advancing faster than society’s ability to test and control them. It wants slower deployment, independent evaluations and international coordination before more powerful systems are released. A second camp views delay as a danger because economic, scientific and security advantages could pass to competitors.
Between them are those who reject a research freeze but favor regulation proportionate to risk. Under this approach, an assistant that drafts an email would not be governed like a system that controls critical infrastructure, supports biological weapons design or makes autonomous decisions affecting fundamental rights. The split shows that the central issue is not only how fast AI moves, but who remains accountable when that power causes harm.
When the builders themselves sound the alarm
Anthropic chief executive Dario Amodei has repeatedly emphasized the possibility that advanced AI could assist cyberattacks or the design of dangerous biological agents. Sam Altman and Elon Musk have also spoken publicly about severe risks, even as they lead companies investing heavily in the race. The changes and nuances in their positions expose a difficult paradox: the people closest to the technology’s potential also face the strongest commercial pressure to keep advancing.

















