Short answer: In October 2026, attackers went after AI tools and the keys that pay for them. Malicious packages, a fake coding extension and breached cloud platforms were all used to steal API keys for services like OpenAI and Anthropic. You can check your own exposure in about 15 minutes: find your keys, rotate any that might be exposed, and lock down which tools can see them.
Summary
Security firm Wiz called it “Black October”: a wave of attacks on AI developer tools and platforms.
Several attacks specifically targeted AI provider API keys, the codes that let software use services like OpenAI and Anthropic, and bill you for it.
A stolen AI key can mean surprise bills, leaked data and attackers using your account.
The fix is basic hygiene: know where your keys live, rotate them, set spending limits and vet the tools you install.
The 15-minute checklist below covers the essentials.
In this article
What happened in “Black October”?
Why are attackers going after AI API keys?
What can happen if your AI key is stolen?
The 15-minute AI key security check
How do you protect AI keys long term?
Common mistakes that expose AI keys
FAQ
What happened in “Black October”?
“Black October” is the name security firm Wiz gave to a wave of attacks on AI tools and developer platforms in October 2026. Not every incident is linked, and AI wasn’t proven to be involved in each one. But many of them shared a target: developer credentials and AI keys.
Here are the incidents most relevant to companies using AI (Anadolu Agency timeline, citing company statements and Wiz, October 2026):
Date (2026) | What happened | Why it matters |
|---|---|---|
8 October | The “GhostAction” campaign added malicious workflows to GitHub repositories through compromised maintainer accounts | It targeted keys for OpenAI, Anthropic and OpenRouter |
16–17 October | OpenCode said about 1.4% of its Console users were affected | Some users reported unauthorized charges on their AI accounts |
20–22 October | Attackers accessed Railway’s internal systems and customer environment variables | Early findings suggest they were hunting for AI provider API keys |
21 October | A malicious version of LiteLLM (1.107.0), a popular open-source AI library, was published with credential-stealing code | Teams that updated automatically may have installed it |
29 October | A tampered Claude Code extension sat in Cursor’s marketplace for about 11 hours | Cursor said up to 5,000 users may have been affected |
Why are attackers going after AI API keys?
Attackers go after AI API keys because the keys are valuable and often poorly protected. An API key is a code that lets software talk to a service. For AI services, the key is also a credit card: whoever holds it can run models on your account.
Free AI access. Attackers can resell stolen access or use it to train their own models. Researchers linked some October incidents to “model distillation”, copying a powerful model by feeding it huge numbers of questions.
Keys are everywhere. Keys end up in code, chat histories, config files and shared documents. AI coding tools make this worse, because people paste keys into prompts to get things working.
Developer tools have broad access. A coding extension or library often runs with the same permissions as the developer, so one bad install can read everything.
What can happen if your AI key is stolen?
A stolen AI key can cost you money, data and trust. The impact depends on what the key can reach.
Risk | What it looks like |
|---|---|
Surprise bills | Thousands of dollars of AI usage you didn’t run, often within hours |
Data exposure | If the key connects to stored files or chat history, those can leak |
Account suspension | Providers may suspend accounts that show abuse, taking your product offline |
Reputation damage | Customers hear that your AI features were part of a breach |
Using AI in your product and not sure where your keys live? Type B Digital runs short security reviews for AI-powered products. Book a 30-minute call and you’ll leave with a prioritized fix list.
The 15-minute AI key security check
You can run this check in about 15 minutes. It won’t make you bulletproof, but it closes the most common gaps.
List every AI key (3 minutes). Open the dashboards for each AI provider you use, such as OpenAI, Anthropic and Google. Note every active key and what it’s used for.
Delete keys nobody uses (2 minutes). Any key without a clear owner or purpose should go.
Rotate anything that might be exposed (3 minutes). If a key was ever pasted in a chat, committed to code, or stored on Railway, LiteLLM or a GitHub workflow during October, create a new key and revoke the old one.
Set spending limits and alerts (2 minutes). Most providers let you cap monthly spend and email you when usage spikes. Turn both on.
Check your installed extensions (3 minutes). Review the extensions in your code editor, such as VS Code or Cursor. Remove anything you don’t recognize or don’t use, and only keep extensions from verified publishers.
Pin your AI library versions (2 minutes). If you use LiteLLM or similar packages, confirm you’re on a safe version and stop automatic updates for critical libraries.
How do you protect AI keys long term?
You protect AI keys long term by treating them like passwords to your bank account. These habits make the biggest difference:
Store keys in a secrets manager, not in code or shared docs. Your cloud platform’s built-in secrets settings are a good start.
Use one key per app or environment, so you can revoke one without breaking everything.
Give keys the least access possible. If a key only needs to run one model, don’t let it do more.
Scan your code for secrets before every release with tools like TruffleHog or GitHub’s secret scanning.
Review new AI tools before installing them, especially extensions and plugins that run inside your editor.
Log and monitor usage, so a spike at 3am triggers an alert, not a surprise invoice.
Common mistakes that expose AI keys
These are the mistakes we see most often in AI-powered products:
Pasting keys into AI chats to “just get it working”.
Putting keys in frontend code, where anyone can read them in the browser.
One master key for everything, so a single leak exposes every system.
Never rotating keys, so an old leak stays useful forever.
No spending cap, so the first sign of a breach is the bill.
FAQ
What is an AI API key? An AI API key is a secret code that lets software use an AI service such as OpenAI or Anthropic. The key identifies your account, so anyone who has it can run AI models and be billed to you.
What was Black October? Black October is the name security firm Wiz gave to a wave of attacks on AI tools and developer platforms in October 2026. Incidents included a malicious LiteLLM release, a tampered Claude Code extension in Cursor’s marketplace and a breach at Railway.
How do I know if my AI key was stolen? Check your AI provider’s usage dashboard for activity you don’t recognize, such as spikes at unusual times or models you don’t use. If you see anything odd, revoke the key immediately and create a new one.
How often should I rotate AI API keys? Rotate AI API keys at least every 90 days, and immediately if a key may have been exposed. Use separate keys for each app so rotation doesn’t break everything.
Are AI coding extensions safe? Most AI coding extensions are safe, but October 2026 showed that fake versions can appear in marketplaces. Install only from verified publishers, remove extensions you don’t use and watch for unusual updates.
Was LiteLLM hacked? A malicious version of LiteLLM, 1.107.0, was published on 21 October 2026 with credential-stealing code. The breach was traced to an exposed admin credential. If you installed that version, rotate every key on that machine.
Secure the keys, keep the AI
AI tools aren’t the problem. Weak key handling is. A few habits, such as secret managers, spending caps, separate keys and careful installs, remove most of the risk.
Type B Digital is a technology consultancy that builds and secures AI-powered products, from security reviews to managed IT and cyber security support. Book a 30-minute call. You’ll leave with a prioritized list of fixes for your AI setup.
Sources
Anadolu Agency: Timeline of the October 2026 cyberattack wave (October 2026, citing company statements and Wiz)
Engadget: 5 safety practices to know before vibe coding (1 November 2026)
