OpenCode is currently under a sustained cyberattack, and mid-incident they ran into the same problem Hugging Face did: the frontier models they were using for defense began refusing to help fight attacking agents because analyzing the attacks started triggering safety guardrails.
OpenCode says frontier models refused to help defend against attacks
AISummary
OpenCode says it is under a sustained cyberattack and that its frontier models began refusing to help analyze the attacking agents because the work triggered safety guardrails. Andrew Curran says Hugging Face hit the same problem. OpenCode says it now has access to Daybreak Blue, which works better for the investigation.
Post on XView on X
Andrew CurranVerified on X
@AndrewCurran_
@RhysSullivan yes that was a bit frustrating mid incident when trying to investigate. they did quite a bit and then started refusing we got access to daybreak blue now so it's betterView quoted post on X
Source: Andrew Curran · x.comPublished · added here