His comments come as alarm over AI safety has intensified in the United States, where leading developers and researchers have warned that increasingly autonomous systems could bypass safeguards or become difficult to control.