Hackers hit AI coding tool OpenCode, running up charges on users’ AI accounts

Your stored API keys in AI developer tools are now a higher risk for unauthorized charges and breaches.

Hackers breached OpenCode's online platform, OpenCode Console, compromising API keys for 1.4% of its users. The attackers exploited a "distillation attack," leading to unauthorized charges on users' AI accounts with providers like OpenAI, Anthropic, and Google. Developers reported unauthorized spending ranging from hundreds to over a thousand dollars. Anthropic and Google blocked affected keys, while OpenAI assisted in refunds. OpenCode is working on refunds for affected users and stated that its open-source app and command-line tool were not compromised. The incident highlights challenges in defending against AI-driven cyberattacks, especially when AI models used for defense have safety filters that can inadvertently hinder incident response. The company has since gained access to OpenAI's Daybreak Blue program, offering AI models with fewer restrictions for security analysis.

The breach of OpenCode Console underscores a critical vulnerability in the burgeoning AI development ecosystem: the security of API keys stored by third-party tools. As developers increasingly rely on these platforms to manage their interactions with multiple AI models, the centralized storage of credentials becomes a prime target for sophisticated, AI-powered attacks. This incident signals a potential arms race where attackers leverage AI to probe for weaknesses and exploit them at scale, while defenders struggle to keep pace using AI tools that themselves can be hampered by built-in safety protocols.

This event raises significant questions about the security posture of AI development platforms and the inherent risks of managing sensitive API keys outside of the direct control of AI providers. It suggests a market where robust, AI-native security solutions will become a key differentiator, potentially driving consolidation or demanding higher security standards from all players. Users must now weigh the convenience of integrated developer tools against the increased risk of compromise, and AI providers may need to offer more granular controls and enhanced monitoring for API key usage.