-
OpenCode says a breach of a legacy system may have exposed AI provider keys saved in OpenCode Console.
-
The incident has impacted about 1.4 percent of Console users that connected to any of OpenAI, Anthropic, or Google providers.
-
Victims are advised to revoke and substitute API keys, check their billing, and set limits on their spendings.
OpenCode says an intruder reached a legacy system on October 16. The company says the incident may have exposed third-party AI provider keys saved in OpenCode Console.
The affected group makes up 1.4 percent of Console users. OpenCode says all affected users had linked an outside provider, such as OpenAI, Anthropic, or Google.
What Happened
OpenCode Console is an optional hosted service. A workspace can save an OpenAI, Anthropic, or Google key in Console. Console then uses that key to call the AI provider. Team members can use the models without seeing the key. OpenCode says the intruder reached a legacy system on the night of October 16. The company says there is a possible compromise of the stored keys in Console.
OpenCode sent warning emails on the morning of October 17. The email subject line was ‘Action Required: Potential provider key breach.’ The company says the incident did not affect users who did not receive the email. It also says the breach involved OpenCode Console, not the OpenCode command-line or desktop apps.
The command-line and desktop apps store provider keys locally in a SQLite database. OpenCode says those keys never reach its servers.
Why the keys matter
A provider key is a bearer credential. Anyone who holds a copy can send billable requests to the account of the owner. The attacker will have access to the key and still use it until the provider revokes it. Deleting the key from Console does not revoke it.
A Console budget also cannot stop unauthorized use. Console only sees requests that pass through its own gateway. OpenCode says it has not confirmed that attackers copied or used any key. However, it warns that exposure creates a serious risk.
Some users say they have already seen unauthorized charges. One user reported $281 in OpenAI usage and $168 in Anthropic usage. Another user reported losses of $1,700. Other users reported $400 in Anthropic usage and about €200 in Google usage.
What affected users should do
Affected users should revoke their provider keys immediately. They should do this inside the OpenAI, Anthropic, or Google dashboard. They should then create new keys. They should update every app, script, and workflow that used the old key.
Users should also review usage and billing records. They should check for requests they did not make, starting from October 16. OpenCode has not said when the unauthorized access began. Therefore, users should review earlier billing records as well.
Users should set spend limits and billing alerts at each provider. They should also give each tool its own API key. This limits damage if one key leaks. It also makes it easier to trace unusual usage to a specific tool.
Why the incident matters
OpenCode says the breach relates to a large-scale distillation attack. Distillation means pulling large amounts of model output to train another AI model.
Anthropic said in September that some groups use fake accounts, stolen cards, stolen login details, and harvested API keys to reach AI models. Criminal markets also resell AI tokens and API keys. Attackers can use stolen keys to run up large bills on a victim’s account.
A separate incident involving digital trust saw hackers hijack three country-code domain registries to obtain Google certificates, highlighting how weaknesses in internet infrastructure can undermine security.
OpenCode says it is working with Anthropic, OpenAI, and Google to fix the issue. The company has not named the attacker or explained how the legacy system was reached.
According to co-founder Dax Raad, the assaults against OpenCode systems commenced on October 18. In his words, attackers have been continuously targeting OpenCode for the last 48 hours. Further, Raad revealed that the attacks compelled OpenCode to harden its systems, but the company hasn’t revealed the technicalities of its security measures to the public.
The event makes it obvious that there should be utmost caution when saving API keys. People should not keep their service provider keys in public systems unless they operate in secure environments. In addition, they should change their keys regularly. Users should also monitor their accounts closely, curtail keyword permissions, & use separate keys for different tools.
