Unlock AI-powered security with Defender for Cloud and DeepSeek R1. Detect identity risks, data exposure, and cyber threats in real-time.
A solid security foundation is the first step towards a successful AI revolution. Given the speed at which Artificial Intelligence is being developed and used, businesses want insight into the new AI tools and applications they are using. To safeguard Artificial Intelligence applications you create and utilise, Microsoft Security offers threat prevention, posture management, data security, compliance, and governance. These features may also assist businesses obtain insight and control over the use of the independent DeepSeek consumer app, as well as protect and manage AI apps developed using the DeepSeek R1 paradigm.
Secure and govern AI apps built with the DeepSeek R1 model on Azure AI Foundry and GitHub
Develop with trustworthy AI
Microsoft Azure revealed DeepSeek R1’s availability on GitHub and Azure AI Foundry, adding to a wide range of more than 1,800 models in it portfolio.
Azure AI Foundry is being used by customers to create production-ready AI applications while taking into consideration their various security, privacy, and safety needs. DeepSeek R1 has passed stringent red teaming and safety tests, including automated assessments of model behaviour and comprehensive security checks to reduce possible vulnerabilities, just like other models offered in Azure AI Foundry. Microsoft’s AI model hosting protections are intended to ensure that client data stays inside Azure’s safe bounds.
In order to assist identify and stop dangerous, malicious, or ungrounded information, Azure AI information Safety comes with built-in content screening by default. Opt-out options are offered for further flexibility. Customers may also effectively test their apps before to deployment with the help of the safety evaluation system. These protections enable Azure AI Foundry to offer businesses a safe, legal, and accountable environment in which to develop and implement AI solutions with confidence.
Start with Security Posture Management
Particularly when developers use open-source materials, AI workloads provide additional threat surfaces and vulnerabilities. In order to identify all AI inventory, including models, orchestrators, grounding data sources, and the direct and indirect risks associated with these components, it is imperative that security posture management be the first step. Defender for Cloud AI security posture management features can assist security teams in gaining insight into AI workloads, identifying AI cyberattack surfaces and vulnerabilities, identifying cyberattack paths that malicious actors could exploit, and receiving recommendations to proactively fortify their security posture against cyberthreats when developers create AI workloads using DeepSeek R1 or other AI models.
Defender for Cloud continuously reveals contextualised security issues and makes risk-based security recommendations that are suited to prioritise critical gaps across your AI workloads by mapping out AI workloads and synthesising security insights like identity risks, sensitive data, and internet exposure. Additionally, the Azure portal’s Azure AI resource itself contains pertinent security advice. This lets developers or workload owners address cyberthreats more quickly by giving them immediate access to advice.
Safeguard DeepSeek R1 AI workloads with cyberthreat protection
A robust security posture lowers the likelihood of cyberattacks, but because AI is dynamic and sophisticated, it also need active monitoring throughout operation. Every AI model may be attacked by cyberthreats like prompt injection. AI application security requires monitoring the latest models.
Defender for Cloud, which is integrated with Azure AI Foundry, continually scans your DeepSeek AI apps for odd or dangerous activity, correlates results, and adds context to security alerts. This gives your security operations centre (SOC) analysts warnings about current cyberthreats, including sensitive data spills, credential theft, and jailbreak intrusions. Azure AI Content Safety prompt shields, for instance, may instantly stop a prompt injection cyberattack. After that, the incident is enhanced with Microsoft Threat Intelligence in Defender for Cloud, which gives SOC analysts insight into user behaviour and supporting data like IP address, model deployment information, and questionable user prompts that caused the alert.
In order to comprehend the whole extent of a cyberattack, including malicious behaviours connected to their generative AI applications, security professionals may centralise AI workload notifications into associated events with these alerts’ integration with Microsoft Defender XDR.
Secure and govern the use of the DeepSeek app
Along with the DeepSeek R1 model, DeepSeek also offers a consumer app that is housed on its local servers. As is frequently the case with consumer-focused apps, the data gathering and cybersecurity procedures may not meet your organisational needs. This emphasises the dangers that companies face if partners and staff use unapproved AI applications, which might result in data breaches and regulatory infractions. Microsoft Security has the ability to identify how third-party AI apps are being used in your company and offers safeguards and guidelines for their use.
Secure and gain visibility into DeepSeek app usage
More than 850 Generative AI apps have ready-to-use risk evaluations available through Defender for Cloud Apps, and the list is updated often as new apps gain popularity. This implies that you may determine how these generative AI apps, like the DeepSeek app, are used in your company, evaluate the security, legal, and compliance issues associated with them, and adjust controls appropriately.Security teams may, for instance, mark high-risk AI apps as unsanctioned and completely prevent users from using them.
Comprehensive data security
Additionally, Microsoft Purview Data Security Posture Management (DSPM) for AI offers insight into data security and compliance threats, including non-compliant usage and sensitive data in user prompts, and suggests countermeasures. For instance, data security teams may develop and improve their data security policies to safeguard sensitive data and stop data breaches by using the reports in DSPM for AI to get insight into the kinds of sensitive data being copied to Generative AI consumer apps, such as the DeepSeek consumer app.
Prevent sensitive data leaks and exfiltration
The leakage of corporate data is one of the primary concerns of security executives about the use of AI. This highlights the need of businesses putting protections in place to prevent users from divulging private information to third-party AI apps.
You may prevent users from copying or uploading private information into Generative AI apps from compatible browsers by using Microsoft Purview Data Loss Prevention (DLP). Additionally, your DLP policy may adjust to the amount of insider danger by imposing more severe limitations on individuals classified as “elevated risk” and less severe restrictions on those classified as “low-risk.”
For instance, while low-risk users are free to carry on with their work, elevated-risk users are prohibited from copying critical data into AI programs. By utilising these features, you can protect your private information from the dangers associated with utilising third-party AI apps. Within Purview, security administrators may then look into these data security threats and conduct insider risk investigations. For comprehensive investigations, Defender XDR presents these same data security threats.
This is a brief summary of some of the features that can assist you in protecting and managing AI applications that you create on GitHub and Azure AI Foundry, as well as AI applications that are used by users inside your company.
