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

Attackers breached an older system belonging to open-source AI coding agent OpenCode, compromising digital passwords and running up hundreds of dollars in unauthorized charges for some users. The weekend incident affected 1.4% of users on the company’s online platform, OpenCode Console, where developers store the API keys used to access external AI services.

Developers took to social media to report sudden spending alerts from OpenAI, Anthropic, and Google. While Anthropic and Google blocked the exposed keys, OpenAI worked with OpenCode to restore access and refund affected customers. Co-founder Dax Raad stated that the command-line tool and core open-source app were unaffected.

OpenCode stated that refunds are underway and instructed impacted users to replace their keys directly with each provider. During the response, OpenCode faced a technical hurdle when defensive AI models refused to analyze malicious code because the payloads tripped safety filters. The company later gained access to OpenAI’s Daybreak Blue program, which provides vetted users with models less likely to block defensive security work.

The incident highlights ongoing challenges faced by organizations trying to defend against automated cyberattacks while navigating strict AI model safety guardrails.