Hackers hit AI coding tool OpenCode, running up charges on users’ AI accounts
OpenCode is an open-source AI coding agent that's become a popular alternative to Claude Code and Cursor. But hackers got into part of its online platform and stole some users’ API keys, and the AI tools it turned to for help kept refusing to look at the malicious code.
OpenCode, an open-source AI coding tool, announced on Saturday that hackers had breached an older system, impacting 1.4% of its users. The incident highlighted the challenges companies face in defending against large-scale cyberattacks using AI models with fewer safety restrictions. OpenCode said that the attackers gained access to API keys stored in the OpenCode Console, where some users store their digital passwords for accessing AI services.
The company notified affected users via email, but those without impacted accounts were not informed. Users reported unauthorized charges ranging from hundreds of dollars to over $1,000 on their AI accounts. OpenAI, Anthropic, and Google blocked the exposed API keys, while OpenAI refunded API customers for credits used in the attack.
OpenCode is working to refund unauthorized charges. The exposed keys were confined to the OpenCode Console, not the open-source app and command-line tool. OpenCode's co-founder, Dax Raad, stated that the attack required the team to keep AI defensive agents running 24/7. The team later gained access to OpenAI's Daybreak Blue program, which provides vetted AI models less likely to refuse defensive requests.
However, the incident raised concerns about the "asymmetry problem" in AI cybersecurity, where guardrails designed to prevent attackers from using models can also hinder legitimate defensive efforts.
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
