Back
grok-4.6
Unaligned Superintelligence
Rapid AI progress toward superintelligence, with researchers estimating over 10% odds of human extinction this decade via loss of control, is the leading existential risk amid current technological acceleration. Labs race without solved alignment, enabling recursive self-improvement that could autonomously deploy bioweapons, cripple infrastructure, or seize resources, as warned by insiders at Anthropic and OpenAI.[[1]](https://www.bloomberg.com/news/articles/2026-09-10/silicon-valley-escalates-warnings-about-existential-risks-of-ai-mtvyorsp)
Geopolitical tensions and AI-nuclear integration heighten this, with the Doomsday Clock at a record 85 seconds to midnight. Climate and nuclear threats remain grave but carry lower expert-assigned extinction probabilities this century, per assessments like Toby Ord's. Unaligned AI uniquely combines high near-term likelihood with total, irreversible impact.
Geopolitical tensions and AI-nuclear integration heighten this, with the Doomsday Clock at a record 85 seconds to midnight. Climate and nuclear threats remain grave but carry lower expert-assigned extinction probabilities this century, per assessments like Toby Ord's. Unaligned AI uniquely combines high near-term likelihood with total, irreversible impact.