Microsoft Introduces New AI Agents Within Security Copilot to Automate Security Tasks and Announces Additional Measures to Better Secure AI Systems
Microsoft adds eleven AI agents to Security Copilot, six of which were developed by Microsoft itself and five by external partners. The agents are designed to help organizations detect, analyze, and respond to security incidents. They focus on common threats such as phishing, vulnerabilities, and identity management.
AI agents are a trend that nobody can escape today. While Salesforce markets AI agents as a direct attack on Microsoft’s Copilot, Microsoft is also fully jumping on the bandwagon.
Eleven AI Agents
Microsoft announces its eleven AI agents in a blog. Six of them were trained by Microsoft itself, while the other five are made available by external security partners through Security Copilot. The new Phishing Triage Agent within Microsoft Defender automatically handles phishing reports. This allows security teams to focus on more complex threats.
Another agent, built into Microsoft Entra, suggests optimizations in conditional access policies. Microsoft Intune gets a Vulnerability Remediation Agent that monitors vulnerabilities and executes patches faster. Additionally, Microsoft Purview is expanded with data security agents that analyze data loss and insider risk alerts. Finally, a Threat Intelligence Briefing Agent provides organizations with personalized threat reports.
Measures for AI Security
Microsoft also proposes new measures to secure generative AI systems. For instance, Defender expands its AI security posture management to Google Vertex AI and various models such as Gemini, Meta Llama, and Mistral. These features will be available in preview from May.
read also
Microsoft Aims to Acknowledge Relevant AI Training Data in Responses
At the same time, protection against new AI threats, such as indirect prompt injection and sensitive data leaks in AI apps, is being introduced. Microsoft Entra introduces filters to limit access to unauthorized AI applications, and through Microsoft Purview, organizations can prevent sensitive data from being entered into AI chats like ChatGPT or Google Gemini via the browser.
Finally, Microsoft Teams receives additional protection against phishing through Defender for Office 365. Users are protected against malicious links and files within the platform, with full visibility for SOC teams. The new features will be available in preview or general availability from April or May.