What happened
Northflank, a cloud application hosting platform, disclosed a security incident on October 30 after receiving a threat alert from security firm Wiz. The company instructed customers to assume all environment variables had been decrypted and to immediately rotate all stored secrets, including AI API keys. Northflank confirmed that multiple customers reported unauthorized usage of their AI API keys stored on the platform. The company applied network policy patches within hours of the alert and enabled audit logs for all users, including those on free plans, to help investigate suspicious activity. However, Northflank has not disclosed the specific attack vector, the total number of affected customers, or whether other types of credentials were compromised.
Northflank disclosed a security incident on October 30, 2026, after receiving a threat alert from cloud security company Wiz at 11:43 UTC. The company's engineers reviewed the alert one minute later and quickly implemented a network policy patch, followed by additional patches at 12:50 UTC and 13:05 UTC. Northflank also requested assistance from Wiz's incident response team.
In its public updates, Northflank warned customers to assume that all environment variables stored on the platform had been decrypted and that the secrets within them were stolen. The company explicitly stated that users had reported unauthorized use of AI API keys supplied to Northflank. Customers were advised to immediately rotate all user-provided secrets, including those in workloads, secret groups, and global secrets, and to disable public access to add-ons where enabled.
As part of its remediation efforts, Northflank enabled credential rotation for PostgreSQL, MongoDB, MySQL, and Redis. The company also enabled audit logs for all customers, including those using free accounts, to facilitate the investigation of suspicious activity. Northflank advised customers to review their AI provider usage and billing logs for unexpected activity and to report any suspicious usage to both Northflank and their respective AI providers.
The company has not disclosed how the attackers gained access to its systems, the total number of customers affected, or whether credentials other than AI API keys were stolen or abused. BleepingComputer contacted Northflank with questions regarding the breach's impact and whether stored credentials were accessed, but the company had not responded as of the report's publication.
Source details: bleepingcomputer.com
Why it matters
This incident highlights the growing security risks associated with cloud platforms that host AI applications, which often store sensitive API keys for expensive AI services. Attackers are increasingly targeting these keys through a technique known as LLMJacking, where stolen credentials are used to generate massive API charges or resell access to AI models. Wiz, which identified this Northflank breach as part of a broader pattern of attacks on AI development platforms, reported that similar incidents have led to significant financial losses, including approximately $100,000 in unauthorized Amazon Bedrock usage in other cases. The breach underscores the critical need for robust credential management and monitoring in AI infrastructure, as compromised keys can lead to direct financial damage and potential data exposure for downstream services.
The breach is part of a broader trend of attacks targeting AI development platforms, which security researchers identify as high-value targets because they store API keys for expensive AI services. Wiz published a report titled 'Black October for AI Developer Platforms,' identifying at least nine security incidents affecting AI development platforms during October, including Northflank, Railway, OpenCode, Warp, and Lightning AI.
Attackers are specifically targeting environment variables containing AI service credentials to exploit a practice known as LLMJacking. This involves using stolen credentials to access AI models via another organization's API keys, generating massive API charges or allowing the stolen access to be resold. In two cases investigated by Wiz, compromised enterprise credentials were used to make approximately 2.5 million Amazon Bedrock API calls, resulting in approximately $100,000 in unauthorized AI usage.
Wiz also discovered infrastructure linked to services reselling access to AI models, although it could not definitively link those operations to the initial breaches or the large-scale abuse. While Wiz has not confirmed that all nine incidents were conducted by the same attackers, many involved attempts to steal AI API keys. Northflank referenced Wiz's research in its incident updates, noting that customer reports of unauthorized AI key usage may be tied to these broader attacks.
The incident highlights a critical vulnerability in AI infrastructure: once attackers gain valid credentials, prevention scores drop sharply. The breach underscores the need for developers and organizations to implement rigorous credential management practices, including frequent rotation and detailed audit logging, to mitigate the financial and security risks associated with AI API key exposure.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Why can ethical evaluation not be reduced to one model score?
What to watch next
Monitor for further disclosures from Northflank regarding the scope of the breach and the specific attack vector. Watch for additional reports of unauthorized AI API usage from other customers or platforms mentioned in Wiz's 'Black October' report, such as Railway, OpenCode, and Warp. Track whether security firms identify common attack patterns or threat actors across these multiple incidents. Also observe if AI providers implement new safeguards or billing alerts to detect and prevent LLMJacking attacks on their platforms.
Monitor for further official disclosures from Northflank regarding the specific attack vector, the total number of affected customers, and whether any data other than API keys was compromised. The company's lack of response to BleepingComputer's inquiries suggests that detailed forensic findings may still be pending.
Watch for additional reports of unauthorized AI API usage from other platforms mentioned in Wiz's 'Black October' report, such as Railway, OpenCode, and Warp. The Railway incident also involved potential exposure of decrypted customer environment variables, while OpenCode disclosed exposed AI API keys, indicating a widespread vulnerability in this sector.
Track whether security firms identify common attack patterns or specific threat actors across the multiple incidents reported in October. Wiz has not confirmed that all incidents were conducted by the same attackers, but the similarity in targeting AI API keys suggests coordinated or opportunistic activity by multiple groups.
Observe if AI providers such as Amazon Bedrock, OpenAI, or Anthropic implement new safeguards, billing alerts, or detection mechanisms to help customers identify and prevent LLMJacking attacks. The financial impact of unauthorized API usage, as seen in the $100,000 loss in other cases, may drive providers to enhance their security offerings.