OpenCode suffered a breach exposing customer API keys for OpenAI, Anthropic and Google. Attackers targeted its systems for 48 hours straight. Frontier AI models initially helped investigate but then refused requests. OpenCode gained access to OpenAI Daybreak Blue for fewer restrictions. OpenAI, Google and Anthropic revoked exposed keys and addressed charges. Hugging Face used a Chinese open-weight model after hitting frontier model limits. Security teams now face a choice between restricted proprietary models and open alternatives. AI companies defend restrictions as necessary to prevent abuse. Cybersecurity teams must balance speed and safety in AI-powered defenses.
What changed
Frontier AI models assisted investigations without restrictions.
What it unlocks
Access to OpenAI Daybreak Blue for vetted cybersecurity defenders.
- 1.4% of OpenCode Console users affected
- 48 hours of continuous attacks
What you need to act on it
- vetted cybersecurity defender status
Sources