AIRO launches automated catastrophic AI risk forecasts, matching top human forecasters
soumitrashukla9 · x · 2026-09-12
The Forecasting Research team released AIRO (Automated AI Risk Outlook), a dashboard of automated catastrophic-risk forecasts that tracks what frontier models predict over time.
- Authors argue frontier models already forecast as well as good human forecasters and will soon be better
- Forecasts are conditional on Epoch AI's Capabilities Index and span 6-month to 2100 windows
- Incident-counting rules were refreshed on each elicitation date; full historical data is downloadable
More from Safety
- Gary Marcus amplifies take: AI development is an unregulated gain-of-function experiment — GaryMarcus · 2026-09-12
- Bipartisan Senate AI Safety Talks Hit Snag as Cantwell Pushes Back on Weak Federal Standard — ShakeelHashim · 2026-09-12
- 'We actually saved a kid': schools deploy AI chatbots amid counselor shortage — deepdishpizza_ · 2026-09-12
- Stanford Used LLMs to Scan 9,623 Local Legal Codes, Finding Dozens Still Mandating Segregation — chrmanning · 2026-09-12
- Meta's Confidential Computing Promise Doesn't Protect Data if Inference Isn't Attested — signulll · 2026-09-12
- Bernie Sanders' AI bill would sentence 'advanced AI' developers and researchers to 20 years in prison — venturetwins · 2026-09-12