Datadog Cloud Security Report: Serious Vulnerabilities Up 4x in Nine Months
Datadog's State of Cloud Security 2026 report finds that the median number of high or critical vulnerabilities in cloud services rose fourfold, from four in February 2026 to around 16 in November. At the same time, more than one in five organisations now use cloud AI services — and long-lived API keys give a stolen credential a long useful life.
Key numbers from the report
- 4 → ~16 — median high or critical vulnerabilities in cloud services (Feb → Nov 2026), a fourfold rise in nine months
- More than 1 in 5 — organisations using cloud AI services
- Nearly half — of organisations using Google Gemini that use long-lived Gemini API keys
- 17% — of organisations using Amazon Bedrock that use long-lived Bedrock keys
- 12 organisations — whose cloud environments showed similar Bedrock-access probing in a 30-day period
- 149 — software packages the average cloud workload depends on, more than half of them indirect dependencies
- 55% — of organisations using AWS that have adopted data perimeters, up from 40% in 2025
What the report finds
- The question is not only that flaws are multiplying, but how far an attacker can get through one compromised application. 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.
- Long-lived AI keys are a new gap: to make AI services easier to adopt, cloud providers offer API keys that remain valid until revoked. “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.
- Attackers are already testing: Datadog Security Labs uncovered tools designed to test whether stolen AWS credentials can access Amazon Bedrock, and researchers observed similar probing in 12 organisations' cloud environments over a 30-day period.
- Supply-chain speed cuts both ways: 60% of organisations run cloud applications using a package version released less than 24 hours earlier, and 21% of cloud applications use a version this new. A compromised version can reach cloud applications fast. Shai-Hulud 2.0 malware shows the consequences: it searched infected systems for cloud credentials.
- There is progress: 55% of organisations using AWS have adopted data perimeters, up from 40% in 2025 — limiting what an attacker can reach even with a valid credential.
What Datadog says
“In the AI era, security becomes an observability problem: you can't stop a system-wide attack without seeing the whole system.” — Emilio Escobar, Chief Information Security Officer, Datadog
What it means for Hong Kong SMEs
For Hong Kong businesses moving to the cloud and using AI, the easiest thing to overlook is key management: once a long-lived API key leaks, an attacker can use it for years. Switching to temporary credentials, rotating regularly and narrowing permissions is the lowest-cost, highest-impact improvement. FinityAI's SRAA checks these foundations, and the free quick check can immediately show your website's basic security posture.
Sources
This article is a FinityAI summary and commentary on a public press release, and is not affiliated with Datadog. Figures come from Datadog's report, based on its customer sample.
Related services
SRAA services · AI Agent security readiness · AI Agent workflow planning