GovAI found no universal jailbreak after 1,000 red-teaming hours.
@krishnanrohit Isn't human break-glass access a common thing? As long as it is monitored and logged, it is as safe a ZDR (because a bad actor might always try to get around it and inspect the network traffic or similar)
Wait, how will *any* enterprise use Fable or Mythos if this is the case?
Positive users praise Anthropic's scalable refusal technology and call the 30-day prompt retention policy fair and reasonable, while some note it creates problems for legacy enterprises.
anthropic has made the situation pretty clear. they've built scalable and robust refusal + monitoring technology, and frontier intelligence is available only under the condition it's used well
this is really a pretty bad time for slow-moving legacy enterprises and also startups selling to legacy slow-moving enterprises. but we kind of already knew that.
They're setting up trusted-access programs for cyber + bio. Basically, a safeguarded model for everyone (Fable 5) – downgraded to Opus 4.8 if classifiers fire – the full-capability version for vetted users (Mythos 5).
They've also clearly put work into vision, where Claude has historically been poor. Fable 5 rebuilt a web app's source code from screenshots alone, and beat Pokémon FireRed with a minimal vision-only harness.
@tenobrus wait why? because they won't adopt mythos because of it's data retention policy?
@tenobrus Downright fair and reasonable behavior tbh.
I wonder what their plan is for space colonization (assuming they're even thinking that far ahead) - do they give everyone a copy of Mythos 10 that kills them if they try to use it to RKV the Earth or something?
GovAI found no universal jailbreak after 1,000 red-teaming hours.
@krishnanrohit Isn't human break-glass access a common thing? As long as it is monitored and logged, it is as safe a ZDR (because a bad actor might always try to get around it and inspect the network traffic or similar)
Wait, how will *any* enterprise use Fable or Mythos if this is the case?