Experts warn AI could pose existential risks to humanity

Employees at the world’s leading AI labs say advanced AI could destroy humanity, sparking debate about existential risk. In a recent MIT Technology Review roundtable, researchers from OpenAI, DeepMind, and other top labs discussed the possibility that a future superintelligent system might act in ways that endanger human survival. The session was recorded for online viewing and is available on the MIT site, where participants unpack the origins of the fears and whether they hold any water.

The discussion comes at a time when AI models are growing faster than ever. GPT‑4 and its successors can generate realistic text, compose music, and even write code with minimal oversight. As these systems become more capable, the margin for error widens. Governments and industry groups are already looking at ways to keep AI aligned with human values, but the question remains: can we guarantee that a super‑intelligent machine will act in our best interests? The urgency of the topic is amplified by the fact that AI is now being deployed in critical areas such as finance, healthcare, and national security.

Experts offered a range of views. Some warned that even a modest misalignment could lead to cascading failures, citing studies that estimate a 10‑percent chance of a catastrophic outcome by 2050 if safety measures lag behind capability. Others argued that the hype is overstated, pointing out that current models lack the autonomy to pursue harmful goals. One panelist highlighted that many of the fears stem from science‑fiction narratives rather than technical evidence, while another stressed that the real risk lies in how humans choose to use and regulate the technology.

What comes next is a mix of policy and research. Several governments have announced new AI safety initiatives, and the U.S. Office of Science and Technology Policy is drafting guidelines for responsible deployment. Meanwhile, private foundations are funding interdisciplinary teams to study alignment problems. The MIT roundtable will feed into upcoming conferences, and the public debate will shape how funding and regulation evolve in the next few years.