🔴 DEVELOPING: OpenCode says the incident is ongoing and has shared no further technical details. Everything below reflects the disclosure as of October 17, 2026.
OpenCode Data Breach: your keys may fund the attack
On October 17, 2026, the open-source AI coding platform OpenCode disclosed "unauthorized access to a legacy system" — one that was being decommissioned. Third-party provider keys (OpenAI, Anthropic and others) saved in users' workspaces may have been exposed. OpenCode calls it a "large scale distillation attack" hitting about 1.4% of Console users — those with external providers linked. Attackers weren't after passwords: they used the keys to hammer high-end AI models, and victims watched their balances drain — some into the hundreds of dollars.
Short answer: if you got OpenCode's email, revoke and replace your provider keys now, update every app using them, and audit your usage and billing. If you didn't get the email, OpenCode says you were not impacted. And this is a real-world distillation attack — the exact threat class we cover in our distillation defense check.
The 60-second OpenCode exposure check
Answer for yourself or your team.
Share your result:
What happened (as disclosed)
October 17, 2026: OpenCode discloses via its X account and individual emails — "unauthorized access to a legacy system" that was being decommissioned; "third party provider keys (eg OpenAI, Anthropic) saved in your workspace may have been exposed."
Scale: about 1.4% of OpenCode Console users — specifically those with external providers linked.
OpenCode's framing: a "large scale distillation attack" — stolen keys used to run repeated queries against high-end models.
Impact: victims' provider balances drained — reported accounts include ~$400 in unauthorized Anthropic charges, ~€200 in Google usage (one project suspended for abuse), and a $1,700 bill. These are user reports, not OpenCode figures.
Remediation: OpenCode says it is working with Anthropic, OpenAI, and Google. Anthropic has revoked affected Claude API keys. As of Oct 18, no provider has said it will refund fraudulent usage — the "who pays?" question remains unanswered.
Who was affected: per OpenCode, if you did NOT receive the email, you were not impacted.
Latest (Oct 18): the "not your keys, not your tokens" self-custody lesson holds — OpenCode confirms keys added to the local CLI via /connect were never part of this incident; only keys synced to Console workspaces were exposed. Timeline: breach overnight into Oct 17, user emails that morning, public statement 17:56 UTC. Developers are treating this as the wake-up call: keep provider keys on your own machine, never parked in anyone's console.
Sources: OpenCode's X disclosure and user email notices, as reported by wpnews.pro and 7minai.com. Claims attributed as stated by OpenCode; details are thin while the incident is ongoing.
What to do NOW
1. Revoke the keys. In each provider's dashboard (OpenAI Platform → API keys, Anthropic Console → API keys, Google AI Studio), revoke every key that was ever saved in an OpenCode workspace.
2. Issue fresh keys. Create replacements and update every app, script, and CI job that used the old ones — a revoked key left in a config breaks the app; a live key left in a config breaks your budget.
3. Audit usage and billing. Review recent usage for spikes you don't recognize — unexpected model calls or regions are the fingerprint of key abuse.
4. Set billing alerts. Hard caps and spend alerts turn the next key compromise into a notification instead of a surprise bill.
5. Scope keys least-privilege. Separate keys per project, restricted to the models and endpoints each project actually needs.
The bigger lesson
Your bill is an attack surface. AI coding tools sit between you and paid APIs — a key compromise is both a security incident and a surprise invoice. This is the first major breach where the theft was the compute.
Decommissioned systems are classic weak points. The breached system was legacy and on its way out — exactly the infrastructure teams stop patching and watching.
Distillation is the motive, not just the method. Attackers don't need your data when your key buys them unlimited queries against frontier models. See our distillation defense check for the full threat class.
FROM INGENIOUS AI
API keys in AI tools are a breach waiting to happen. BreachHunter watches for exposure so a leaked key becomes an alert, not a surprise bill.
On October 17, 2026, OpenCode disclosed unauthorized access to a legacy system that was being decommissioned. Third-party provider keys (e.g. OpenAI, Anthropic) of about 1.4% of Console users were exposed in what the company called a large-scale distillation attack.
Was I affected by the OpenCode breach?
Per OpenCode: if you did NOT receive their email notice, you were not impacted. Affected users received individual email notices with guidance to revoke and replace their keys.
What should I do if my API keys were exposed?
Revoke and replace the keys via each provider's dashboard (OpenAI, Anthropic, Google), update every app using the old keys, and review recent usage and billing for unfamiliar activity.
What is a distillation attack?
An attack that uses stolen API access to run repeated queries against a powerful model — harvesting its outputs to train or inform a cheaper copy, or simply to resell the access.
What is a data breach via API key exposure?
A data breach via API key exposure happens when attackers obtain the secret keys that let applications call paid AI services — as in the OpenCode incident, where a legacy system leak exposed provider keys. Whoever holds the keys can run up bills and harvest model outputs at the victim's expense.
Were your keys in the OpenCode data breach?
If you used OpenCode Console and did not get their email notice, you were not affected — only about 1.4% of users were. If you did get the notice, revoke every exposed key at the provider dashboard today; keys cannot be "unseen" once leaked.