How did the founder end up with $449 in charges?

Stolen API keys were used to run up charges on the founder's OpenAI and Anthropic accounts. The owner found out only after the bill accrued.

How do stolen API keys typically get exposed?

Keys are commonly committed to code repositories, sit in environment variables within compromised cloud environments, or are copied into tools, scripts, or developer machines that later get breached.

Why can an attacker profit from an API key without touching the victim's app or data?

An API key works like a password for software, letting whoever holds it make requests billed to the account that owns the key. The attacker simply uses it for AI model calls at the victim's expense, so the key itself is the product.

Black October Stolen API Keys Cost Founder $449 in Hours

Q: What is Black October?

Black October is a name used for a cluster of cyberattacks tied to cloud systems and developer tooling, with AI playing a role in how quickly the attacks move. The article notes it should be treated as shorthand for a pattern rather than a confirmed single campaign.

Q: Why can't the dollar amount be used to judge the risk?

The $449 figure is modest but reflects how little time and effort the abuse requires. Without spending caps, the same access could run up a far larger bill.

A startup founder woke up to $449 in unexpected OpenAI and Anthropic charges. The cause, according to Fox News, was a set of stolen API keys exploited during the so-called "Black October" cyber wave. The amount is small next to the damage a large company could face, but the story shows how Black October stolen API keys can turn into real bills before anyone is awake to notice.

This post covers what the label refers to, how stolen keys tend to be abused, why AI-assisted attacks leave defenders less time, and what developers can do today.

What 'Black October' Refers To (and What Is Not Yet Confirmed)

"Black October" is the name being used for a cluster of cyberattacks tied to cloud systems and developer tooling, with AI playing a role in how quickly the attacks move. The Fox News report centers on one concrete consequence: credentials taken from cloud environments and code repositories were used to run up charges on AI services.

It is worth being careful about what is and is not established. The article gives us one documented example, the founder's $449 in charges on OpenAI and Anthropic accounts. It does not, in the material we reviewed, provide a full technical account of every incident under the Black October banner, nor does it establish a single attacker or a single method behind them. Other coverage appearing in search results describes multiple AI developer platforms disclosing breaches in October and malicious software packages built to steal secrets, but those details vary by outlet and we have not independently verified them. Treat the label as shorthand for a pattern, not a confirmed single campaign.

How Stolen API Keys Are Exploited Within Hours

An API key is a long string of characters that works like a password for software. Whoever holds it can make requests as if they were you, and the provider bills the account that owns the key. There is usually no username prompt and no second factor at that point, which is why a leaked key is so useful to an attacker.

Keys end up exposed in a few common ways:

They are committed to a code repository, sometimes in a configuration file or an old commit.
They sit in environment variables or settings inside a cloud environment that gets compromised.
They are copied into tools, scripts or developer machines that later get breached.

Once an attacker has a working key, the simplest move is to use it for AI model calls at the victim's expense. That fits the founder's experience: the charges appeared on OpenAI and Anthropic accounts, and the owner found out only after the bill accrued. The attacker does not need to touch the victim's app or data to profit. The key itself is the product.

This is also why the dollar figure can mislead. $449 is modest, but it reflects how little time and effort the abuse requires. Without spending caps, the same access could run up a far larger bill.

Why AI-Assisted Attacks Shrink Defenders' Response Time

The "AI" in AI-assisted attacks matters because of speed. Searching repositories for secrets, testing whether a key is valid, and deciding what to do next are tasks that automation handles well. When those steps are chained together, the gap between a leak and abuse can drop from days to hours.

We covered a related development in Sysdig's finding of what it describes as a fully autonomous AI ransomware attack, where an AI agent carried out an intrusion with little or no human direction. That is a different type of incident from API key theft, but the underlying lesson is similar: automated attackers do not wait for business hours, and manual response processes can fall behind.

For small teams the problem is sharper. A startup rarely has a security operations center watching billing dashboards overnight. The founder in this story found out by checking charges after the fact, which is how many people discover this kind of abuse.

What This Means For You

If you build with AI APIs, even on a side project, your keys are valuable credentials and should be handled like passwords to a payment account. Three points are worth keeping in mind:

Exposure is often invisible. You may not know a key has leaked until the charges arrive.
Your provider bills you. Stolen usage generally shows up on your account, so prevention and early detection are on you.
Small bills are a warning. An unexpected $449 charge means a key worked for someone else. Treat it as a security incident, not just an accounting problem.

How to Protect API Keys: Rotation, Scoping, Alerts, MFA and Secrets Scanning

None of these steps requires a security team. Most take less than an hour.

Rotate keys now. Generate new keys and revoke the old ones, especially any that have appeared in code, chat, tickets or shared documents. Rotate again after any suspected exposure.
Scope each key narrowly. Use separate keys per project or environment, and give each only the permissions it needs, if your provider supports it. A leaked test key should not unlock production.
Set spending limits and alerts. Configure budget caps and usage notifications in each provider's billing settings so unusual activity triggers a message instead of a surprise.
Turn on multi-factor authentication. Protect the provider accounts and cloud consoles where keys are created and viewed. MFA will not stop a stolen key from working, but it makes it harder to steal keys from the account itself.
Scan for secrets. Run a secrets scanner on your repositories, including commit history, and check CI settings and shared config files. Remove any exposed keys and rotate them, since deleting a file does not undo a leak.
Keep keys out of code. Store them in a dedicated secrets manager or protected environment variables, not in source files.

Takeaways

The Black October stolen API keys story is a reminder that small, quiet credentials can cost real money fast. If you build with AI APIs, audit and rotate your keys today, set spending limits and alerts, and scan your repositories for exposed secrets. For broader context on how autonomous AI attacks are speeding up intrusions, read our report on Sysdig's autonomous AI ransomware findings, then spend an hour tightening your own setup.