Evan Hubinger Alignment Science Lead at Anthropic warns of the potential threat of AI.

[https://www.theguardian.com/technology/2026/sep/09/ai-superintelligence-risks-warnings-scientists-politicians](https://www.theguardian.com/technology/2026/sep/09/ai-superintelligence-risks-warnings-scientists-politicians)

Hubinger admitted that the industry currently lacks a plan to solve the AI alignment problem.

* He believes there is a greater than 10% probability of catastrophic harm/extinction within a decade due to rapid self-improvement in AI models.
* Former Anthropic researcher Jacob Coxon recently resigned, calling out major labs for engaging in an irresponsible race toward super-intelligence.
* Anthropic expects AI self-improvement in R&D to hit critical safety threshold levels within 6 to 12 months.

Main risks mentioned include autonomous bioweapon creation, loss of human control over supere-intelligent systems, and large-scale cyber warfare disabling power/water grids.