A security researcher has identified 26,400 valid cloud and artificial intelligence credentials exposed in public Gitee repositories after scanning millions of publicly available code repositories for leaked secrets. Independent researcher Ben Zimmermann discovered the credentials by analyzing 7.25 million public repositories hosted on Gitee, a Chinese code hosting platform, and published the findings on October 28. The exposed credentials included access keys and tokens linked to various cloud and AI services, creating potential security risks for organizations and developers whose secrets were publicly accessible.
According to the research findings, the largest group of exposed credentials involved Alibaba Cloud AccessKey pairs, with 9,473 active keys identified during testing. Other exposed credentials included 3,805 YouTube API keys, 3,170 DeepSeek keys, 1,489 keys associated with Zhipu’s BigModel AI platform, and 1,348 OpenAI keys. Smaller numbers of exposed credentials were also found for services including GitHub, AWS, Google Cloud, and Google Gemini. Zimmermann provided vendors with lists of credentials associated with their services but did not directly contact developers who had uploaded the keys. At the time of publication, he had not confirmed whether all reported credentials had been revoked, meaning developers who previously exposed keys should not assume those credentials are no longer accessible. The research involved scanning repositories over approximately one month using two secret detection tools, TruffleHog and Kingfisher. Both tools tested discovered credentials against the services that issued them to determine whether they were still active. The scanning process included repository commit histories, meaning credentials that had been removed from visible code could still be identified if they remained available in previous commits. The researcher noted that the 26,400 figure represents unique credentials counted once, regardless of how many repositories contained the same secret. Based on commit dates, 12,242 exposed credentials originated from 2025, while some credentials dated back to earlier years.
The findings also highlighted risks associated with exposed credentials, as active keys can allow unauthorized use of cloud resources, API services, or account quotas depending on the permissions attached to them. For example, compromised OpenAI API keys can potentially result in unexpected usage charges, exhausted quotas, or disruption of API access. Similarly, improperly restricted Google Cloud API keys may be used with services beyond their intended purpose. Security experts recommend that developers immediately revoke and replace any credentials that have appeared in public repositories, as removing keys from code or rewriting repository history does not invalidate previously exposed secrets. Zimmermann advised organizations to rotate exposed credentials through their respective service providers and review activity logs for suspicious usage. For Alibaba Cloud users, recommended steps include creating new AccessKey pairs, updating applications, disabling old keys, and reviewing activity records. Google and YouTube API users should rotate affected API keys, update applications, and apply restrictions to limit usage. OpenAI users should rotate exposed API keys immediately and configure spending controls to reduce potential impact from unauthorized usage.
The research also compared results from TruffleHog and Kingfisher, showing differences between the two scanning tools. TruffleHog identified 12,737 working credentials, while Kingfisher detected 14,664. Only 1,001 credentials were identified by both tools, showing that different scanners may produce different results depending on their detection capabilities. The findings highlight the importance of using multiple security checks when monitoring repositories for exposed secrets. The discovery comes amid wider efforts across open source ecosystems to improve software supply chain security. Similar measures, including publishing delays and additional security checks, have been introduced across platforms and package ecosystems to reduce the risk of newly released malicious or compromised components spreading through automated development workflows. The Gitee findings further emphasize the need for developers and organizations to maintain strong secret management practices and avoid storing sensitive credentials in publicly accessible repositories.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
