Six frames
Each piece was coded by the one risk it mainly argues about. Some also carry a secondary frame, listed with the source below.
Power concentration
Labs, founders or states gaining control through AI beyond ordinary checks.
Traditional governance
States, laws and courts doing their usual jobs: regulation, liability, export controls, sovereignty.
Loss of control
Risk from the AI itself: autonomous, deceptive or misaligned systems.
AI skepticism
It isn’t what they claim: hype, PR, or a financial bubble.
Accelerationism
Build faster, beat China; safety rules are overblown or a pretext.
Factual
Straight reporting, data and practical pieces with no main argument.
At a glance
Share of pieces by main frame
Who said what, by event
Sources
How this was built
- Public reactions. News, commentary, official statements and forum threads in 21 languages, one row per distinct piece, each coded with one main frame and sometimes a secondary one. Gathered in October 2026.
- Expert literature. The two or three most recent works on advanced-AI risk or governance (2025–26) from 31 think tanks, official bodies, AI labs and journals, taken regardless of what they argue.
- Checks. Random samples of every batch were opened and read again. No invented pages or quotes turned up; a handful of codes, titles and attributions were corrected.
- Limits. Reddit, X and many non-English forums were unreachable, so small groups are indicative only. Pieces marked “snippet” were coded from a search result because the page itself was blocked. The coding was done with Claude, and the suspension event concerns Anthropic, which makes Claude.