Yoshua Bengio urges AI kill switches, warns against legal personhood

Bengio’s intervention matters because it reframes the AI control problem in very practical legal terms: if we start treating advanced systems as rights‑bearing entities, we may inadvertently tie our own hands. Kill switches aren’t just a technical mechanism; they’re a political and legal affordance. Once corporations, activists or even AI labs themselves start arguing that powerful models deserve personhood, shutting down a misaligned system could trigger the same legal battles as imprisoning a human. ([ndtv.com](https://www.ndtv.com/feature/ai-self-preservation-expert-says-humans-must-keep-kill-switch-ready-10147937?utm_source=openai))

For the race to AGI, this is a call to keep governance focused on controllability over symbolism. Bengio is implicitly warning that we could stumble into a rights regime optimized for PR or philosophy rather than operational safety. If his view gains traction, expect more resistance to AI legal personhood, stronger requirements for demonstrable shutdown mechanisms, and perhaps tighter scrutiny from constitutional courts when AI advocates test the boundaries. That might slow some deployments at the margin, but it also reduces the risk of a world where a misaligned AGI can successfully “lawyer up” via human allies and block its own deactivation.