Anthropic's J-Lens research indicates so. Like when web results are fake it woul... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

user43928 14 days ago | parent | context | favorite | on: GPT-6 Astra

Anthropic's J-Lens research indicates so.

Like when web results are fake it would show things like 'FAKE PROMPT INJECTION'.

Or in a safety evaluation with a contrived scenario it was 'FAKE FICTIONAL'.

qsera 14 days ago [–]

Sure. That is easy.

But I was talking about training data.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact