Evidence item

AI Frontiers reports that major labs keep models internal for months before public release during testing and refinement. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 10366

Source: The Hidden AI Frontier · locator {} · excerpt hash sha256:6a76105feb609d33bc3abcdb39bc0670994d2240c2157ce5da6ed397bd4e031e

Each of the models that power the major AI systems you’ve heard of — ChatGPT, Claude, Gemini — spends months as an internal model before public release. During this period, these systems undergo safety testing, capability evaluation, and refinement.