“That is why self-replication is widely recognised as one of the few red line risks of frontier AI systems.” AI safety has become an increasingly prominent issue for researchers and ...