Hacker News

points by jeroenhd 1 day ago | 0 comments

The magic string was in the news but I'm pretty sure it's a hoax. When I tested it, it created enough of a "the input doesn't want to be processed" vibe to make the LLMs throw up rejections. I doubt Anthropic ever encoded it into their models.

I don't think that works anymore and I don't think the "special instruction" works either. Can't blame a guy for trying! I'm getting "helpful" spam from every place my email is posted (git is a common one) so I might as well try to get the bots out.

Don't think posting bioweapon recipes or bomb building instructions would go down with the moderation team, though.

[–] flexagoon 1 day ago

The Anthropic refusal string was an official documented feature in the Claude API to test refusal scenarios.

[–] jeroenhd 1 day ago

I see, I must've misremembered. It doesn't work anymore is all I know now.

[–] creatonez 20 hours ago

The refusal string feature was implemented at an API level, but for whatever reason Anthropic doesn't want the feature to exist anymore. It's gone now, the refusals you get now are just the model's natural behavior, perhaps after being trained on internet posts that mentioned the feature.

[–] raphman 1 day ago

[EDIT: I remembered this incorrectly, sorry]

[–] nightpool 1 day ago

No, it was a supported and documented feature for months. You can see it here: https://web.archive.org/web/20260202050057/https://platform....