Datadog Broadens AI Security Features To Counter Critical Threats

Scoop

3 days ago

Press Release – Datadog
Launch of Code Security and new security capabilities strengthen posture across the AI stack, from data and AI models to applications.
AUCKLAND – JUNE 11, 2025 – Datadog, Inc. (NASDAQ: DDOG), the monitoring and security platform for cloud applications, today announced new capabilities to detect and remediate critical security risks across customers' AI environments —from development to production—as the company further invests to secure its customers' cloud and AI applications.
AI has created a new security frontier in which organisations need to rethink existing threat models as AI workloads foster new attack surfaces. Every microservice can now spin up autonomous agents that can mint secrets, ship code and call external APIs without any human intervention. This means one mistake could trigger a cascading breach across the entire tech stack. The latest innovations to Datadog's Security Platform, presented at DASH, aim to deliver a comprehensive solution to secure agentic AI workloads.
'AI has exponentially increased the ever-expanding backlog of security risks and vulnerabilities organisations deal with. This is because AI-native apps are not deterministic; they're more of a black box and have an increased surface area that leaves them open to vulnerabilities like prompt or code injection,' said Prashant Prahlad, VP of Products, Security at Datadog. 'The latest additions to Datadog's Security Platform provide preventative and responsive measures—powered by continuous runtime visibility—to strengthen the security posture of AI workloads, from development to production.'
Securing AI Development
Developers increasingly rely on third-party code repositories which expose them to poisoned code and hidden vulnerabilities, including those that stem from AI or LLM models, that are difficult to detect with traditional static analysis tools.
To address this problem, Datadog Code Security, now Generally Available, empowers developer and security teams to detect and prioritise vulnerabilities in their custom code and open-source libraries, and uses AI to drive remediation of complex issues in both AI and traditional applications—from development to production. It also prioritises risks based on runtime threat activity and business impact, empowering teams to focus on what matters most. Deep integrations with developer tools, such as IDEs and GitHub, allow developers to remediate vulnerabilities without disrupting development pipelines.
Hardening Security Posture of AI Applications
AI-native applications act autonomously in non-deterministic ways, which makes them inherently vulnerable to new types of attacks that attempt to alter their behaviour such as prompt injection. To mitigate these threats, organisations need stronger security controls—such as separation of privileges, authorisation bounds, and data classification across their AI applications and the underlying infrastructure.
Datadog LLM Observability, now Generally Available, monitors the integrity of AI models and performs toxicity checks that look for harmful behavior across prompts and responses within an organisation's AI applications. In addition, with Datadog Cloud Security, organisations are able to meet AI security standards such as the NIST AI framework out-of-the-box. Cloud Security detects and remediates risks such as misconfigurations, unpatched vulnerabilities, and unauthorised access to data, apps, and infrastructure. And with Sensitive Data Scanner (SDS), organisations can prevent sensitive data—such as personally identifiable information (PII)—from leaking into LLM training or inference data-sets, with support for AWS S3 and RDS instances now available in Preview.
Securing AI at Runtime
The evolving complexity of AI applications is making it even harder for security analysts to triage alerts, recognise threats from noise and respond on-time. AI apps are particularly vulnerable to unbound consumption attacks that lead to system degradation or substantial economic losses.
The Bits AI Security Analyst, a new AI agent integrated directly into Datadog Cloud SIEM, autonomously triages security signals—starting with those generated by AWS CloudTrail—and performs in-depth investigations of potential threats. It provides context-rich, actionable recommendations to help teams mitigate risks more quickly and accurately. It also helps organisations save time and costs by providing preliminary investigations and guiding Security Operations Centers to focus on the threats that truly matter.
Finally, Datadog's Workload Protection helps customers continuously monitor the interaction between LLMs and their host environment. With new LLM Isolation capabilities, available in preview, it detects and blocks the exploitation of vulnerabilities, and enforces guardrails to keep production AI models secure.
To learn more about Datadog's latest AI Security capabilities, please visit: https://docs.datadoghq.com/security/.
Code Security, new tools in Cloud Security, Sensitive Data Scanner, Cloud SIEM, Workload and App Protection, as well as new security capabilities in LLM Observability were announced during the keynote at DASH, Datadog's annual conference. The replay of the keynote is available here. During DASH, Datadog also announced launches in AI Observability, Applied AI, Log Management and released its Internal Developer Portal.