Archived report

2026-07-25: Advisory

Original forecast
Advisory forecast
Evidence available at the time
Anthropic: 1 open incident: Elevated errors for Claude Fable 5, Claude Sonnet 5, Claude Haiku 4.5, and other models; OpenAI: ChatGPT Health opens to all US adult users
Later outcome
Directionally correct
Forecast grade
directional
What changed afterward
Issued as Advisory; grading after the window closes.

← Back to the history archive