Experts rank dangerous AI capabilities and cyberattacks as the top catastrophic risks
gamersecret2 · reddit · 2026-07-28
An Axios piece highlights five AI risks that experts see as having the highest potential for catastrophic harm.
The chart in the post summarizes a late-2025 survey of 272 international experts. The top concerns are AI systems acquiring dangerous capabilities, cyberattacks and weaponized use, power concentration and unfair distribution of benefits, competitive dynamics, and false or misleading information. The post frames these as the risk areas experts most worry could lead to large-scale harm between 2025 and 2030 if current trends continue.
This is less about a specific policy proposal and more about the broader AGI-era risk landscape: what kinds of failure modes are keeping researchers and policy people up at night.
More from AGI Musings
- AI’s real path was less theory and more data, compute, and gradient descent — burny_tech · 2026-07-28
- Poster says no one serious is still working on deep learning or reinforcement learning — fkasummer · 2026-07-28
- LLM mania could still delay AI diffusion by rattling the economy — fkasummer · 2026-07-28
- AI-for-math has hit short-horizon wins, but the next leap may need deeper theory — burny_tech · 2026-07-28
- AI should not be steered by a handful of companies, the post argues — ctjlewis · 2026-07-28
- AI will become more probabilistic and adaptive, and ethics will matter more — AryHHAry · 2026-07-28