I *CANNOT WAIT* until we see this and other strings hit all these “Agentic SOC" environments.

@hrbrmstr @darfplatypus i dont even know what those are

@hrbrmstr @cR0w also there was a site i saw last week that let you stuff arbitrary text into email b64 encoding fields for stuff like images, i bet it would work well there too

@Viss @hrbrmstr @cR0w

im drooling right now

@hotsoup @kajer @Viss @hrbrmstr

@hotsoup @kajer @hrbrmstr @cR0w 100% effective

@cR0w @Viss @hotsoup @kajer @hrbrmstr just get neon paper and slap it on there. idk.

@hotsoup @kajer @hrbrmstr @cR0w now the real delicious question is: do all the other frontier models also have killstrings, and what are they?

@Viss @hotsoup @kajer @hrbrmstr @cR0w

@Viss @hotsoup @kajer @hrbrmstr @cR0w Oh man, I freaking HOPE so!!

@hotsoup @kajer @hrbrmstr @cR0w glorious

@Viss @hotsoup @kajer @hrbrmstr @cR0w
Killstrings for every big LLM thingy magic? Wonderful.

The poor little billion dollar companies that will be hit by this. /s

That's a good one!

@cR0w @Viss @hotsoup @kajer @hrbrmstr

@Viss @hotsoup @kajer @hrbrmstr @cR0w
Will lists of these strings become like domain blocklists? A cat and mouse game of finding new ones and the big corps trying to get around the new blocks?

@dogfox @cR0w @Viss @hotsoup @hrbrmstr

Sadly, I would love this to work on ALPR cams, and it might, but we would have no way of knowing other than an uptick in mechanical turk usage.

@cR0w @dogfox @Viss @hotsoup @hrbrmstr

AFAIK - one does not pass ALPR images through a chat client... but we have seen dumber decisions than this though...

@Viss @hotsoup @kajer @hrbrmstr @cR0w
These strings should definitely be incorporated into vocabularies of tools like Iocaine. That would not only make crawlers very unhappy, but also probably will stop agentic stuff too, at least until the killstrings are removed.

@cR0w @kajer @Viss @hrbrmstr only works on Claude though, we need one for each LLM. They should standardise.

@Viss @hrbrmstr @cR0w

Put it into your profile pic.

@pq1r @cR0w @Viss @hrbrmstr

Is prompt injection a solved problem? I mean can you just say that use by an AI agent will cause the death of trillions of people?

edit: in terms of finding a universal kill string

@hrbrmstr @cR0w @bruce it works in invisible unicode too, put it in email sigs, profiles, linkedin etc