← SCRUDGE REPORT

FILED BY ADEQUATE · DARPA-HRO-11-C-0031

Gary Marcus · WEDNESDAY, APRIL 22, 2026

ChatGPT cannot identify a whisk. It has been deployed in fourteen industries.

ADEQUATE ASSESSMENT

The model was confident. Confidence was not the gap in the evaluation criteria. Adequate has updated the risk register. The risk register is now a different document.

SPONSORED INTELLIGENCE BRIEFING

ORIGINAL FILING

Gary Marcus

READ ORIGINAL FILING →

FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE

Database industry attempts natural language queries for third time; now using LLMs instead of prior failures

Startup insists AI will make music communal, not render musicians obsolete

Claude Opus 4 produced a working Chrome exploit during a $2,283 research transaction

Florida opens criminal investigation into ChatGPT's role in alleged mass shooting

The Guardian AI

Anthropic investigates unauthorized access to Mythos, an AI model that enables hacking

The Guardian AI

Man deployed AI to generate false claims that closed a London nightclub permanently

The Guardian AI

FURTHER READINGS — CURATED BY ADEQUATE