Breach and immediate impact
OpenCode said attackers broke into an older system overnight that affected 1.4% of users of its online platform, OpenCode Console, where some customers store API keys used to access AI services. The company wrote on X that it notified all affected users by email; users who did not receive an email were not impacted, it added.
Developers posted spending alerts and reports of unauthorized charges on social platforms. Some reported individual bills in the hundreds of dollars, and a few described totals above $1,000. OpenCode said refunds for unauthorized charges are underway.
Vendor responses and defensive limits
Anthropic and Google blocked API keys they believed had been exposed. OpenAI said it worked with OpenCode to restore access and refund API customers for credits used in the attack. OpenCode co-founder Dax Raad wrote on X that the exposed keys were confined to the Console and that the open-source app and command-line tool were not affected.
Raad also described sustained intrusion attempts over 48 hours that forced the team to run AI defensive agents continuously. He said some AI models initially helped with analysis but then refused to examine malicious code after safety filters activated. Raad reported that OpenCode later received access to OpenAI’s Daybreak Blue program, which he said improved the team’s ability to investigate the incident.
Context and ongoing actions
OpenCode has not disclosed how attackers gained access to the older system, how keys were stored, or the full extent of misuse. The company has told affected users to revoke and replace their keys directly with each provider. OpenCode said the incident remains ongoing.
Security researchers and industry figures have highlighted a pattern in which some commercial models refuse requests needed for defensive analysis; Hugging Face and David Sacks have previously described similar limits. The breach occurred the same night OpenCode made GLM-5.3-Flash free for users; researchers have noted the broader GLM-5.3 family’s cybersecurity capabilities approach those of restricted-access models like Anthropic’s Claude Mythos Preview.
CrowdStrike and India’s cybersecurity agency recently reported other AI-enabled intrusions, underscoring a series of related incidents. OpenCode continues to coordinate with AI providers and to notify affected users while processing refunds and mitigation steps.
Original source: TechCrunch AI