Datadog tells that cloud faults are four times more than in May

Datadog makes tools for cloud security. It made its cloud security report public on November 10, with data from thousands of customer organizations in the world. Datadog tells that the median number of high or critical faults in a cloud workload went from 4 in May to 14 in June and 16 in November. In May the Anthropic Mythos model came out.

What the report covers

The count comes from third party libraries in applications that Datadog Code Security monitors. The report also looks at keys that do not expire and at new versions that teams use too fast. Datadog tells teams to do these steps:

• Get AI systems with an identity that the workload has for a short time.
• Add an interval before a new dependency version goes live.

The numbers

Almost half of the organizations that use Gemini keep Gemini keys that do not expire. Almost 30% of AWS and Google Cloud users have access keys that were made more than three years before. 60% of organizations operate versions that were made less than 24 hours before.

What is not known

The report is from Datadog, and the source shows no external check. Datadog gives frontier AI as the cause. The source shows no test of that link.

Insight: Three reports show the same weak point, the AI key

Wiz tells that 9 AI developer platform breaches in October got secret keys for AI access. Attackers used each key in 30 minutes, and one key then made 1.7 million calls. At this time Datadog tells that almost half of Gemini users keep keys that do not expire. The key, not the model, is the weak point. Replace static keys first.