October 2026 was a rough month for developers who build with artificial intelligence. Cloud security company Wiz reported on October 30 that at least nine AI developer platforms disclosed security incidents during the month. Wiz called the wave “Black October,” and while the incidents were separate, they shared a worrying pattern: attackers went straight for the stored secrets that let apps talk to AI models.
An AI developer platform is where a developer hosts code and stores the credentials an app needs to run. Those credentials often include API keys for AI services, database passwords, and tokens for other tools. That makes the platform a high-value target. Break into one account store, and an attacker may gain access to many customers’ keys at once.
What happened, platform by platform
Wiz’s list runs from Fireworks AI on October 6 through Northflank on October 30. In between were Andromeda, Zeabur, Lightning AI, OpenCode, Railway, Warp, and CommandCode. The details varied. Fireworks AI said internal credentials were used without authorisation to reach secrets from evaluation jobs. Zeabur reported that millions of environment-variable records were read, and later said it would shut down by December 31. Lightning AI revoked access for all users after finding unauthorised access to credentials. Railway described privilege escalation in its build infrastructure and said up to about 1.85 percent of workspaces were affected.
It is important not to overstate what linked these cases. Wiz was clear that there was no single vulnerability or shared dependency connecting all nine incidents. What they shared was a target and a motive: environment variables, and especially AI keys, which can be used or resold to run expensive AI models at someone else’s expense. Wiz noted that stolen keys were tested within 30 minutes in some cases it studied, and estimated a bill of more than $100,000 from just two keys.
Why stolen AI keys are so costly
A stolen password usually gives access to one account. A stolen AI API key can be different, because every model call it makes is billed to the key owner. Attackers do not need to steal money directly. They can simply burn through inference capacity, resell access, or use the outputs to train a competing model — a practice known as distillation, which OpenCode and CommandCode both mentioned in their disclosures.
That also explains why deleting a key from a platform is not enough. If the key itself is still valid at the provider that issued it, it can still be used elsewhere. Wiz’s central advice was to revoke keys with the issuing provider, not just remove them from the platform where they were stored, and then check usage and billing for calls you cannot explain.
A practical checklist for developers
First, rotate any key that was stored on an affected platform, and revoke it at the provider. Second, replace long-lived, all-purpose keys with short-lived, per-project keys that can do only what one app needs. Third, set spending caps and billing alerts, so unusual usage is caught in hours rather than at the end of the month. Fourth, keep secrets out of code and out of shared logs, and use a proper secrets manager where possible. Finally, review which tools and platforms hold your keys, and remove access you no longer use.
What this means for everyone else
You do not need to be a developer to take a lesson from Black October. The same habit — reusing one powerful credential everywhere — is what makes ordinary accounts easy to compromise too. Use a password manager or passkeys where they are offered, turn on two-factor authentication for email and financial accounts, and treat any alert about unusual sign-ins or spending as a reason to act immediately, not later.
Black October is also a reminder that security is now part of building with AI, not an afterthought. The tools are powerful, but the keys that unlock them deserve the same care as a bank password. Rotate them, limit them, watch them — and if a platform you use reports a breach, assume your key needs replacing until you know otherwise.
Related reading: passkeys in 2026, AI phishing defenses in 2026, AI coding assistants in 2026. External sources: GitHub secret scanning, OWASP secrets management
