Nearly 1,000 readers submitted questions about artificial intelligence to the New York Times, reflecting widespread confusion about the technology's rapid development and potential risks. This month, officials at the two biggest American AI companies celebrated mathematical breakthroughs while simultaneously warning the technology might eventually destroy humanity.

Readers asked how AI might gain agency to act without human control. Currently, AI agents can edit files, send emails, and write code, but lack true agency—humans must direct them and can shut them off. Concerns persist that connecting agents to the internet, power grids, stock markets, or military weapons could cause unexpected havoc, or that researchers might eventually develop systems with genuine agency.

Regarding existential threats, experts cite speculative scenarios: malicious actors using AI to develop bioweapons, systems hoarding economic resources, or AI triggering drone wars or nuclear launches. Guardrails exist at companies like OpenAI, Anthropic, and Google; chatbots typically refuse requests to build bioweapons. On privacy, AI apps like ChatGPT on phones cannot access user folders, passwords, or bank account information. Questions also addressed job displacement, AI costs versus human developers, water consumption by data centers, and why AI outperforms traditional Google searches.