The median number of high- or critical-severity vulnerabilities in cloud services analysed by Datadog increased fourfold, from four in February 2026 to around 16 in November, according to the company’s State of Cloud Security 2026 report. The findings point to growing exposure across cloud applications, alongside risks from persistent AI credentials and software supply chain dependencies.
The report found that more than one in five organisations analysed by Datadog use cloud-based AI services. However, long-lived API keys used to access these services can remain valid until revoked, potentially giving attackers prolonged access if the credentials are compromised. Nearly half of organisations using Google Gemini were found to use long-lived Gemini API keys, while the corresponding figure for Amazon Bedrock was 17%.
“Attackers don’t need to use their own AI model if they can steal the key used to access yours”, said Andrew Krug, Head of Security Advocacy at Datadog. “Teams should use temporary access wherever possible as they build AI applications, rather than leave behind keys that can keep working for years.”
Datadog researchers also observed attempts to test stolen cloud credentials for access to AI services. The company said it identified tools designed to check whether compromised AWS credentials could access Amazon Bedrock, alongside similar probing activity in the cloud environments of 12 organisations over a 30-day period.
Software dependencies present another potential entry point. Cloud workloads analysed in the report relied on an average of 149 software packages, with more than half being indirect dependencies. The report also found that 60% of organisations run cloud applications using package versions released less than 24 hours earlier, increasing the potential exposure to compromised software updates.
The report cited the Shai-Hulud 2.0 malware, which searched infected systems for cloud credentials, as an example of how supply chain attacks can expose access to cloud environments.
There were also signs of progress in limiting the reach of compromised credentials. The share of organisations using AWS that had adopted data perimeters rose to 55%, from 40% in 2025.
“As serious flaws multiply, the question is how far an attacker can get through one compromised application,” said Emilio Escobar, Datadog’s Chief Information Security Officer. “Teams need to fix those flaws and limit the damage if an application is compromised. Temporary credentials and data perimeters can restrict an attacker’s access, while a shared view of applications, identities and infrastructure helps Security and Engineering contain an attack before it spreads. In the AI era, security becomes an observability problem: you can’t stop a system-wide attack without seeing the whole system.”