- A malicious individual has announced he is selling an OpenRouter database containing nearly 250,000 pieces of personal information and 207,847 API keys.
- With stolen API keys, hackers can carry out AI workloads against the victims via a method known as LLMjacking.
- OpenRouter has not verified the breach but all users of OpenRouter should change their API keys as a precaution.
Recently, a malicious hacker has made a shocking announcement in an underground social network. The hacker claims to have access to a database that belongs to OpenRouter, a service that allows developers to work with different AI systems. The post also claims that there are a huge number of API key codes and records that correspond to accounts of users.
OpenRouter provides one API for hundreds of AI models from different providers. Developers use its keys to run AI tools and apps. If the hacker’s announcement is true, criminals could misuse the keys in the victims’ names.
What the Seller Claims to have
The actor lists specific numbers in the forum post. They claim to have 251,996 user rows, which they say covers about 250,000 unique users. They also claim to hold 207,847 API keys. The seller says the keys use the sk-or-v1 format. This matches the key structure of OpenRouter. The actor claims the keys are a mix of active, revoked, and expired.
The alleged user data includes emails and usernames – it also reportedly contains names, signup dates, credits, and plan details. Password material is also mentioned, with some hashed and some claimed as plaintext. Country information is listed as well.
For the API keys, the seller lists user linkage and key values. Creation and last-used timestamps are also claimed. Spend limits and rate limits are part of the alleged dataset. The mix of active and revoked keys matters. Revoked keys no longer work. Active keys still do. If even a small share of the 207,847 keys remain active, that is still a large number of working credentials.
Why Stolen API Keys Matter
API keys work like passwords for developer tools. A stolen key can let someone use an AI service without permission. The bill goes to the account owner.
Attackers can abuse stolen keys to run large AI workloads. This practice is called LLMjacking. It can cost victims money and disrupt their operations. Attackers can also use the keys for other harmful activity.
Some attackers resell access to stolen keys. Others use them to build products of their own. Either way, the victim pays for the computing power. Costs can climb fast when someone runs large models at scale.
OpenRouter has built-in controls for keys. With those, developers are able to set credit limits and decide which models will be accessible through each specific key. Such controls aid in limiting some risks if a key leaks. However, these controls only work if users actually implement them.
The alleged compromise of email and credential details in the claim is also worrisome. Exposed emails can support the launch of phishing campaigns. If the information about passwords in the claim is correct, then the attackers might try to use stolen credentials for access to other services. They may be successful as many people reuse the same passwords across several platforms.
The Claim is Still Unproven
OpenRouter has yet to substantiate any claims that it had a breach. So far, it has not made any announcement on the topic. There is also no competent investigator who has managed to authenticate the alleged information about the breach.
Experts in cybersecurity say that such posts on underground forums always lack proof and are sometimes just a lie. Some sellers like to put exaggerated or made-up information, and in other cases, they use outdated data from previous violations. In this particular case, there were no copies, screenshots or any similar materials available in the post.
The timing also matters. A separate incident in August this year exposed API keys from a cloud platform. That leak included OpenRouter keys among others. Researchers also found valid OpenRouter keys in infostealer logs earlier in the year. Stolen keys have circulated before. The current claim could draw from those older leaks.
Forum sellers often bundle data from several sources. They then present it as one fresh breach. That tactic makes the listing look more valuable than it is. Without samples, researchers cannot tell which parts of this claim are new.
In a separate incident, threat actors also claimed an e-Benefits Prévoyance data breach involving an alleged API flaw, highlighting the need to verify breach claims before treating them as confirmed.
What OpenRouter Users Should Consider
Users should check their API keys and account activity. They should look for unexpected usage or charges. Also, they should discard or replace any possibly compromised key. Simply creating a new key will not automatically disable the old one. Users must revoke the old key manually. This step prevents anyone who has obtained the compromised key from accessing the services.
Users must also check their spending limits and access policies for models. Implementing a daily credit limit helps to minimize losses in the case of any unauthorized use. By limiting the use of keys to one specific model, it will significantly reduce the damage from a stolen key.
Further, making sure that OpenRouter passwords and logins remain secure will also help users a lot. They should avoid using the same password as on other websites. The use of two-factor authentication, if available, provides an additional layer of protection for accounts.
Moreover, developers should also keep an eye on their usage dashboards. If there is a surge in the number of any requests, it could mean that the key has been stolen. Implementing an alert for any suspicious activity can help to catch abusers on time.
Share this article
About the Author
Farwa is an experienced InfoSec writer and cybersecurity journalist skilled in writing articles related to cybersecurity, AI, DevOps, Big Data, Cloud security, VPNs, IAM, and Cloud Computing. Also a contributor on Tripwire.com, Infosecurity Magazine, Security Boulevard, DevOps.com, and CPO Magazine.
More from Farwa SajjadRelated Posts
Thousands of Websites Used to Deliver Coruna and DarkSword iPhone Spyware
Researchers found a campaign called DeadTag that abuses abandoned ad and analytics domains to send v...
PHP Warns of New Potential RCE Flaw, Recommends Disabling Upload Progress
PHP has disclosed CVE-2026-115560, a possible remote code execution flaw that affects apps with an u...
Google and Other Tech Giants Block eMudhra Certificates After Weak Keys Exposed
Researchers cracked eight very weak secret keys tied to Indian certificate company eMudhra within se...
Microsoft Admits Windows 11 October Updates Failed to Fix Always On VPN Bug
Microsoft says its October 2026 Windows 11 updates did not fix the Always On VPN connection bug. The...
Canva Confirms Salesforce Data Exposure via Canny as Hackers Claim Millions of Records
Canva says hackers got into Canny, a tool it used for product feedback, and accessed some business c...
Phonelyne Faces Unverified Data Leak Claim Involving eSIM Codes and API Keys
A hacker called DaOnlySpark claims to hold customer, payment, and eSIM records tied to Phonelyne. Th...
