OpenCode Says AI Defenders Refused Mid-Attack

The models helping OpenCode fight off an attack stopped helping partway through.

OpenCode says AI models that were helping defend against an ongoing cyberattack began refusing further assistance mid-incident, @kimmonismus reports. Access to Daybreak Blue then improved the situation.

Similar problems have reportedly occurred at Hugging Face, he adds, so this does not appear to be unprecedented. The Sand Paper reported OpenCode’s Console breach in Sunday’s edition.

Mollick: keep agents away from your bank account, for now.

Ethan Mollick, separately, says you don’t want to give today’s agents access to control your bank account, “given how little we know about agentic security & prompt injection.”

He argues that finding a secure way to work with agents is probably better for everyone in the long run than a ban.