Daniel Kokotajlo, a former OpenAI researcher and current head of the “AI Futures Project” based in Berkeley, California, is accusing major technology companies-including Anthropic, OpenAI, X, and Google-of engaging in mere “safety washing” regarding the regulation of artificial intelligence. According to Kokotajlo’s remarks to “Der Spiegel”, while these companies pushed for “a little bit of regulation, a few external auditors,” they have largely avoided discussing the necessity of slowing down or halting the automatic development of AI.
Kokotajlo is considered one of the most recognized experts on the future of artificial intelligence and has been warning for years about the technology’s existential risks. He posits that the tech industry is planning to automate AI research, which he warns constitutes humanity crossing a critical threshold. Beyond this point, he suggests that AI could develop superhuman capabilities in various aspects of life-perhaps in every aspect. He calls this development “extremely dangerous.” If an AI begins training and accelerating itself, the pace of progress will increase dramatically. This, according to the researcher, will make it increasingly difficult to monitor the process and more likely that errors will go unnoticed.
Therefore, he is calling for massive government restriction on the AI development conducted by companies like OpenAI and Anthropic. “We essentially need to shut down AI development,” Kokotajlo states. He argues that humanity currently lacks “the understanding and the ability to regulate AI with a scalpel and differentiate between good and bad.” He insists that development of superintelligence should only proceed once a safe and broadly beneficial path has been found. Achieving this, he concludes, requires “complete transparency, a broader distribution of power-and a kill-switch.”
