OpenAI and Anthropic Remedy Security Breach Hitting 1.4% of OpenCode Users

The breach impacted 1.4% of OpenCode Console users who had linked external providers, with some developers reporting losses between $168 and $1,700. During the active response, OpenCode stated that frontier AI models initially refused to assist in analyzing the exploit because the malicious payloads triggered safety guardrails. To overcome these refusals and stabilize its defenses, the company integrated Daybreak Blue, a cyber permissive defense model provided by OpenAI.