OpenCode Console breach may have exposed saved AI provider keys
OpenCode says a security breach on the night of 16 to 17 October 2026 hit 1.4% of OpenCode Console users, a subset with external providers linked. Provider keys saved in a workspace may have been exposed. Affected users were emailed on 17 October. The incident is ongoing. Revoke and reissue linked keys at each provider and check usage.
By ThreatWire Desk · Published 18 Oct 2026
On this page
What happened
OpenCode says it had a security breach on the night of 16 to 17 October 2026. The company announced it on its official X account on 17 October. It said the breach was "related to a large scale distillation attack" and affected 1.4% of OpenCode Console users.
OpenCode says it emailed every affected user on the morning of 17 October. In its words, anyone who did not get an email was not impacted. It says the affected users were a subset who had external providers linked.
The company says the incident is ongoing and it cannot share more details yet. It says it is working with Anthropic, OpenAI and Google to remediate.
Who is affected
The 1.4% figure is about OpenCode Console users. Console is an optional online service. It is not the open source OpenCode agent itself, and the OpenCode docs say Console is completely optional.
Per the OpenCode docs, Console is a pay-as-you-go gateway. Teams can also connect their own provider accounts there. Press reports say Console stores those keys on its servers.
If you only run the open source agent and never signed in to Console, nothing public so far says you are affected. If you use Console and received no email, OpenCode says you were not impacted.
What is confirmed
- OpenCode confirmed the breach and the 1.4% figure on its official X account.
- The notification email is reported by The Latent and Cyera. It is not something ThreatWire has seen directly. Users and security writer Matt Johansen shared it.
- The reported subject is "Action Required: Potential provider key breach".
- The reported text says a third party got unauthorized access to "a legacy system that we were in the process of decommissioning".
- It says provider keys saved in the workspace "may have been exposed". It asks users to revoke them at each provider and watch usage.
- OpenCode says it is working with Anthropic, OpenAI and Google.
What users report, not verified
Several users say they saw charges they did not make. The Latent cites one user who reported $281 of OpenAI usage and $168 on Anthropic. It cites others who reported roughly $700 to $1,700. The Latent says it could not verify these reports. Treat them as unverified accounts.
What is not known
OpenCode has not disclosed:
- the absolute number of affected users
- how the intruder got in
- how the keys were stored
- how the breach connects to the "distillation attack"
- whether it will reimburse users
- whether any data besides provider keys was involved
As of this writing, Anthropic, OpenAI and Google have not commented publicly, according to TSN Media. The attack has no public attribution. ThreatWire does not attribute it to anyone.
Context on distillation
Anthropic described "distillation attacks" on 12 October 2026 as industrial-scale campaigns to extract a model's capabilities, often through fraudulent accounts. Cyera points to an Anthropic report from September that mentions API keys "illegally harvested". Neither source ties these to the OpenCode incident. Whether they are related is not established.
What to do now
If you use OpenCode Console with linked providers:
- Revoke every provider key linked in your Console workspace. Do it in each provider's own dashboard: Anthropic, OpenAI, Google, AWS Bedrock and any other.
- Create new keys and update every app that used the old ones. Do not reuse an old key.
- Check billing and usage at each provider for the last few days. Look for spend or requests you did not make.
- Set spend limits and usage alerts on each provider account.
- Review audit logs where the provider offers them, and keep watching for several weeks.
- Do this even if you did not get an email and use Console with linked keys, if you want to be cautious. OpenCode says no email means not impacted, but revoking costs little.
- Report unauthorized charges to the provider's support, and keep your usage records.
If you only use the open source agent with your own keys, nothing public says those keys were exposed. Keep normal key hygiene.
Sources: OpenCode on X, OpenCode docs, The Latent, Cyera, TSN Media and Notebookcheck. This article will be updated as OpenCode shares more.
Sources
- OpenCode on X, official statement (17 October 2026)
- OpenCode docs, Console (optional pay-as-you-go gateway)
- The Latent, OpenCode says breach tied to distillation attack exposed users' AI provider keys
- Cyera, OpenCode breach, what happened and how to rotate exposed AI API keys
- TSN Media, OpenCode Console breach and distillation
- Notebookcheck, OpenCode hacked, users urged to revoke their API keys