Research Intern - MSRC AI Security Research

The Microsoft Security Response Center (MSRC) works to protect customers and Microsoft from current and emerging threats to security and privacy. Within MSRC, our AI Vulnerabilities and Mitigations team analyzes all reported security vulnerabilities in Microsoft’s AI systems and develops new mitigations through deep research.

We are looking for Research Interns to work with us on developing new mitigations for AI systems. Our team is uniquely placed to solve real-world security and privacy challenges, through cutting-edge scientific research, informed by vulnerability data from production AI systems. Some of our team’s recent research and collaborations include: 

• TaskTracker: Catching LLM Task Drift with Activation Deltas https://arxiv.org/abs/2406.00799
• LLMail-Inject Adaptive Prompt Injection Challenge https://microsoft.github.io/llmail-inject/) 
• Highlight & Summarize https://arxiv.org/abs/2508.02872
• Design Patterns for Securing LLM Agents against Prompt Injections https://arxiv.org/abs/2506.08837 
• Compromising Autonomous LLM Agents Through Malfunction Amplification https://aclanthology.org/2025.emnlp-main.1771/ 
• Securing AI Agents with Information-Flow Control https://arxiv.org/abs/2505.23643
• The Hawthorne Effect in Reasoning Models https://arxiv.org/abs/2505.14617 
• VerifiableFL: Verifiable Claims for Federated Learning using Exclaves https://arxiv.org/abs/2412.10537  
   

• Conduct cutting-edge research in AI security and privacy by proposing, exploring, and evaluating new ideas.
• Develop and implement prototypes to validate concepts and demonstrate real-world applicability.
• Collaborate with cross-functional teams to integrate security and privacy solutions into products and services.
• Publish findings and present insights to internal stakeholders and external conferences to advance industry knowledge. 

Required/Minimum Qualifications:   

• Students enrolled in a PhD program or outstanding undergraduate/master’s students with research experience.  

Preferred/Additional Qualifications: 

• One or more papers at top security conferences (e.g., USENIX, CCS, S&P, NDSS) or papers focusing on security, safety, or privacy and appearing at top machine learning conferences (e.g., NeurIPS, ICML, ICLR) are strongly desired. 
• We’d love to see your work — please share any personal projects, GitHub repositories, portfolios, or examples of work you’re proud of in your application.