OpenCode breach hits 1.4% of Console users; AI defense stalls

OpenCode's breach sits at the intersection of two problems the incident made visible at once. The company said the intrusion hit an older system behind OpenCode Console, where some users stored API keys, and that it affected 1.4% of Console users; the open-source app and command-line tool were untouched, according to co-founder Dax Raad, who stressed that those keys never reach OpenCode itself.

The response exposed a second issue. Raad said attackers tried to break into its systems nonstop for 48 hours, forcing the team to keep AI defensive agents running around the clock. The models it turned to helped for a while, then began refusing to work once the malicious code they were asked to examine tripped safety filters — a wall Hugging Face described in July as the asymmetry problem, writing that guardrails cannot distinguish an incident responder from an attacker. Access to OpenAI's Daybreak Blue program, which gives vetted users models less likely to refuse defensive security work, improved the team's investigation, Raad said.

The breach occurred the same night OpenCode made GLM-5.3-Flash free for users, and it lands amid a run of AI-enabled cyber incidents, including CrowdStrike's report that a China-based hacker likely used an AI penetration-testing agent and Claude in attacks on South Korean banks, and an Indian agency warning that AI agents had reached Indian organizations' systems without authorization.