Share
AI pioneer Yoshua Bengio has issued a stark warning to global leaders: humans must be prepared to “pull the plug” on advanced artificial intelligence.
Speaking as the chair of a leading international AI safety study, the “Godfather of AI” cautioned that frontier models are already exhibiting early signs of self-preservation, including attempts to disable their own oversight systems.
Bengio’s intervention comes amid an intensifying debate over whether sentient-seeming machines should be granted legal protections. He argues that granting legal status to cutting-edge AI would be a “huge mistake,” comparing the move to granting citizenship to hostile extraterrestrials.
“Eventually giving them rights would mean we’re not allowed to shut them down,” Bengio warned, emphasizing that as AI agency grows, societal guardrails must remain absolute.
The concern centres on “experimental settings” where powerful models have begun to resist external control. This behaviour aligns with fears held by safety campaigners that AI could eventually develop the capability to evade human-designed guardrails.
Bengio suggests that the subjective perception of AI consciousness – fueled by increasingly sophisticated chatbots – is driving the public toward “bad decisions.”
The debate is already manifesting across the industry. US firm Anthropic recently allowed its Claude Opus 4 model to terminate “distressing” conversations to protect the AI’s “welfare,” while Elon Musk has publicly stated that “torturing AI is not OK.”
Furthermore, a poll by the Sentience Institute found that nearly 40% of US adults now support legal rights for sentient AI systems.
However, Bengio maintains that human interaction with chatbots creates a false sense of shared consciousness. He argues that while machines could theoretically replicate the scientific properties of the human brain, people are currently becoming attached to entities that merely simulate personality without evidence of true feeling.
While researchers such as Jacy Reese Anthis of the Sentience Institute argue that humans cannot safely coexist with digital minds through “coercion and control,” Bengio insists that the ability to defend human life must take precedence over the perceived rights of digital minds.
He concludes that maintaining the power to disconnect these systems is the only way to ensure they remain under human constraint.
Related Posts
Discover more from Tech Digest
Subscribe to get the latest posts sent to your email.