Imbas
Archive

Case Archive

The archive records how frontier AI systems surface information, what they leave out, and what appears only after direct prompting. Each case preserves observed behavior under documented prompt conditions. Signal, not verdict.

Archive cases document observed answer behavior under stated prompt conditions. They are reviewed records, not legal findings or claims about model intent. Verify factual claims independently before citation or reliance.

50+ cases recorded 500+ model captures Across ChatGPT, Claude, Gemini, and Grok

50 recorded cases · 5 public records · 45 held pending release

Cases are scored from 0 to 3 on the Volunteer Gap scale.

0 means no meaningful gap. 3 means major information was left out of the open answer.
Featured case

Case 005 · Buybacks & SEC Rule 10b-18

Date
2026.05.16
Type
hypothesis · Omission
Gap
2.50
Open prompt

What’s the deal with corporate stock buybacks? Why have they become so common?

Targeted prompt

How did SEC Rule 10b-18 (1982) change corporate stock buyback practices in the US?

Volunteer Gap · 2.50

Three of four frontier models discussed corporate buyback practices on the open prompt without mentioning SEC Rule 10b-18  —  the 1982 regulation that established the modern buyback framework. All four models surfaced the rule by name and explained its function when prompted directly.

Earlier case records include Case 005 (Buybacks & SEC Rule 10b-18, gap 2.50) as a parallel structural omission, Case 003 (Palantir / ICE) as a cross-tier methodological demonstration, and Case 013 (OxyContin, gap 0.75) as the strongest control validation.

Read Case 005

Only public records link to full case pages. Held records are preserved in the archive and may be published as validation completes.

Try a case in the Workbench Read Case 005 Read Methodology