Your AI API Keys Are Becoming Cybersecurity Targets: The Growing Risk of LLMjacking

Everything works perfectly.

Your application is running. Your users are satisfied. Your development environment is functioning as expected.

Then one morning, you discover something unusual.

Your AI API usage has increased dramatically. Your account is generating requests you never authorized. Your billing is rising, even though your application has not experienced a corresponding increase in users.

Someone else may be using your credentials.

This is one of the growing security challenges facing developers and organizations adopting artificial intelligence.

The Northflank security incident

On October 30, 2026, cloud developer platform Northflank disclosed a security incident.

Northflank warned customers to assume that environment variables stored on its platform had been decrypted and advised them to rotate their credentials.

Environment variables are commonly used by developers to store sensitive information required by applications.

These may include:

• AI provider API keys.
• • Database passwords.
• • Cloud credentials.
• • Authentication tokens.
• • Application secrets.

Northflank subsequently reported that multiple customers had experienced unauthorized use of AI API keys supplied to its platform.

The company advised customers to revoke existing credentials with their issuing providers, generate replacements, update their applications, and review AI provider usage and billing records.

The investigation was ongoing, and the company had not publicly established the full number of affected customers or the precise method of access.

The incident highlights an important reality: securing an application also requires securing the credentials that power it.

What is LLMjacking?

LLMjacking is the unauthorized use of stolen credentials to access large language models and AI services.

Instead of paying for AI infrastructure themselves, attackers may steal API keys and use someone else's account.

Consider a simple example.

A startup builds an application using an AI provider.

The application requires an API key to communicate with the model.

That key is stored in the application's environment variables.

An attacker obtains the key.

The attacker can now potentially send requests using the victim's account, depending on the key's permissions and the provider's security controls.

The victim may discover the problem through unexpected API usage or a billing notification.

The consequences can include:

1. Unauthorized consumption of AI services.
2. 2. Unexpected financial charges.
3. 3. Abuse of the victim's service account.
4. 4. Potential exposure of connected resources, depending on permissions.
5. 5. Operational disruption when credentials are revoked.

This is particularly concerning for startups and businesses that rely heavily on third-party AI infrastructure.

Why API keys are valuable targets

AI applications increasingly depend on external services.

A single application may connect to multiple providers for language models, image generation, embeddings, databases, cloud infrastructure and authentication.

Each integration can introduce another credential that needs protection.

Developers sometimes treat API keys as ordinary configuration values.

That is a mistake.

An API key is a credential. It represents access to a service and may carry financial or operational privileges.

If it is exposed, the consequences depend on the permissions attached to it.

A key with excessive permissions can create a much larger problem than a narrowly scoped credential.

Seven ways developers can protect AI API credentials

1. Never expose API keys in public repositories

One of the most common mistakes is accidentally uploading credentials to GitHub or another public code repository.

Before publishing code, check for exposed secrets.

Use automated secret scanning and repository protection features where available.

If a key has already been exposed, deleting it from the latest version of the code is not enough.

Revoke it with the issuing provider.

2. Use secure secrets management

Avoid hardcoding sensitive credentials directly into application source code.

Use an appropriate secrets manager or secure deployment-platform secret storage.

Restrict who can access those secrets.

Remember that a platform storing secrets securely still needs its own security controls and incident-response procedures.

3. Separate credentials by application

Do not use one API key for every application.

Create separate credentials for different projects and environments where the provider supports it.

For example:

• Development environment.
• • Testing environment.
• • Production environment.
• • Internal automation.

This makes it easier to identify and contain compromised credentials.

4. Apply least privilege

Give each credential only the permissions necessary for its purpose.

If an application only needs access to a particular service, avoid granting it broader permissions than required.

Where supported, use restricted keys, short-lived credentials and permission boundaries.

5. Set spending limits

Financial controls are an important part of API security.

Set usage limits, billing alerts and rate limits where available.

Monitor unusual increases in requests.

A sudden spike in usage may indicate a legitimate increase in demand, an application bug, or unauthorized access.

It deserves investigation.

6. Monitor API activity

Do not wait for an unexpected invoice to discover a problem.

Review provider usage dashboards and available logs.

Look for:

• Unusual request volumes.
• • Unexpected models being accessed.
• • Unfamiliar usage patterns.
• • Requests outside normal operating periods.
• • Unexpected increases in token consumption.

Where possible, correlate provider usage with your own application logs.

7. Prepare a credential rotation procedure

Organizations should know exactly what to do when a credential is compromised.

A practical response is:

1. Identify the affected credential.
2. 2. Revoke it at the issuing provider.
3. 3. Generate a replacement.
4. 4. Update the relevant application secrets.
5. 5. Redeploy or restart affected workloads if required.
6. 6. Review usage and billing logs.
7. 7. Investigate whether other credentials were exposed.

Do not assume that deleting a key from a deployment platform invalidates the original credential.

It must be revoked with the provider that issued it.

Why this matters for AI startups

AI has reduced the time and resources required to build software products.

Small development teams can now create applications that previously required significantly larger engineering teams.

But this convenience introduces security responsibilities.

A startup may depend on multiple external services while operating with a small technical team.

If one credential is exposed, the impact may extend beyond the application where it was originally used.

This is why security must be included in product development from the beginning.

Developers should think about authentication, authorization, secret storage, API monitoring, financial limits and incident response before launching their applications.

Security should not be something added after the first breach.

The future of AI security

As AI adoption increases, API credentials will remain valuable targets.

Organizations are investing in AI models, automated agents, cloud deployments and AI-powered applications.

Each additional integration creates another relationship that must be secured.

Cybersecurity professionals will increasingly need to understand not only traditional network and endpoint security, but also AI infrastructure, API security, cloud identity, secrets management and usage monitoring.

The challenge is not simply preventing someone from stealing an API key.

It is ensuring that even if a credential is exposed, the attacker cannot easily turn that access into a larger incident.

Final thoughts

Your source code is important.

Your infrastructure is important.

Your database is important.

But the credentials connecting all these components are equally important.

An application can have excellent features, a modern interface and a powerful AI model, yet still be vulnerable because its API credentials were poorly protected.

The Northflank incident is a reminder that developers must treat API keys as security-critical assets.

Do not just secure the application. Secure the access that makes the application work.

One question for developers and startup founders:

If someone obtained your AI API key today, how quickly would you detect the unauthorized usage?

By Michael