Before letting AI read them, identify invisible characters in PDFs and Word documents. Two scripts to pick up white text and 1pt text, plus a checklist

When you have an AI summarize a PDF, what the AI is reading is not the paper you see on your screen.It is a bundle of text ...
Anthropic

Investigating unintended model actions in our evaluations and internal use

Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.