AI People Wiki
Portrait of Mark Russinovich
Photo: MicrosoftPDC, CC BY 2.0. Credit

Builders & open source

Mark Russinovich

Chief Technology Officer and Technical Fellow, Microsoft Azure, Microsoft

  • Chief Technology Officer and Technical Fellow, Microsoft Azure, Microsoft1

Mark Russinovich is Chief Technology Officer and Technical Fellow for Microsoft Azure. In a 2025 Microsoft Security blog series on the company's deputy CISOs, Russinovich describes a team focused on technical strategy, architecture, and security risk management for the Azure platform, engineering systems, and core operating systems. Russinovich created the Sysinternals web site in 1996 to host advanced system utilities; Microsoft Learn now hosts the Sysinternals tools for troubleshooting Windows and Linux systems. Russinovich is first author of the Crescendo paper on multi-turn jailbreak attacks against language models, accepted at USENIX Security 2025, and co-authored "Who's Harry Potter?" with Ronen Eldan on approximate unlearning in language models.12345

At a glance

Current role
Chief Technology Officer and Technical Fellow, Microsoft Azure
Company
Microsoft
Sector
Company
Education
Carnegie Mellon
Work listed
3
Sources cited
11
Profile reviewed

Career

In 1996, Russinovich created the Sysinternals web site to host advanced system utilities, and later co-founded Winternals Software; Microsoft acquired the company in 2006, when Russinovich joined Microsoft. Microsoft Learn now hosts the Sysinternals tools for managing and troubleshooting Windows and Linux systems.13

Russinovich is Chief Technology Officer and Technical Fellow for Microsoft Azure. In a 2025 Microsoft Security blog series introducing the company's deputy chief information security officers, Russinovich described working with a large team to identify and resolve security risks across Azure, the core operating system and the company's engineering systems.12

Research

Russinovich's earlier co-authored research covers cloud systems, including a characterization of serverless workloads at a large cloud provider (2020), the Singularity scheduler for AI workloads (2022) and the Confidential Consortium Framework for secure multiparty applications (2023).678

More recent work concerns the security of language models. With Ronen Eldan, Russinovich studied approximate unlearning in "Who's Harry Potter?" (2023), and as first author described the Crescendo multi-turn jailbreak (2024) and GRP-Obliteration (2026), which removes a model's safety alignment using a single unlabeled prompt. Russinovich also co-authored lessons from red teaming 100 generative AI products and a design for securing AI agents with information-flow control.5491011

Writing

Russinovich has written nonfiction and fiction books, including the Microsoft Press Windows Internals series, Troubleshooting with the Sysinternals Tools, and the cybersecurity thrillers Zero Day, Trojan Horse and Rogue Code.1

Work and contributions

  • Crescendo multi-turn jailbreak attack Published research

    First author of "Great, Now Write an Article About That: The Crescendo Multi-Turn LLM Jailbreak Attack" (2024, accepted at USENIX Security 2025), which introduces a multi-turn jailbreak that starts with seemingly benign interactions.4

  • Who's Harry Potter? Approximate Unlearning in LLMs Published research

    Co-author with Ronen Eldan of a 2023 paper proposing a technique for unlearning a subset of a language model's training data without retraining from scratch, demonstrated on Llama2-7b.5

  • Sysinternals

    Created the Sysinternals web site in 1996 to host advanced system utilities and technical information; Microsoft Learn now hosts the utilities for managing, troubleshooting, and diagnosing Windows and Linux systems and applications.3

Timeline

  1. Created the Sysinternals web site to host advanced system utilities.3
  2. Joined Microsoft when Microsoft acquired Winternals Software, which Russinovich co-founded.1
  3. Co-authored "Serverless in the Wild", a characterization of serverless workloads at a large cloud provider.6
  4. Co-authored the paper on Singularity, a planet-scale scheduler for AI workloads.7
  5. Co-authored "Who's Harry Potter? Approximate Unlearning in LLMs" with Ronen Eldan.5
  6. Co-authored the Confidential Consortium Framework paper on secure multiparty applications.8
  7. Posted the Crescendo multi-turn jailbreak paper as first author.4
  8. Co-authored "Lessons From Red Teaming 100 Generative AI Products".10
  9. Was profiled in a Microsoft Security blog series introducing the company's deputy chief information security officers.2
  10. Co-authored "Securing AI Agents with Information-Flow Control".11
  11. Posted GRP-Obliteration, on unaligning language models with a single unlabeled prompt, as first author.9

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

11 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Mark Russinovich — Microsoft Azure Blog author pageMicrosoft Azure · Official page · accessed Oct 21, 2026
  2. Meet the Deputy CISOs who help shape Microsoft's approach to cybersecurityMicrosoft Security · Official page · published Apr 8, 2025 · accessed Oct 21, 2026
  3. Sysinternals — Microsoft LearnMicrosoft · Documentation · accessed Oct 21, 2026
  4. Great, Now Write an Article About That: The Crescendo Multi-Turn LLM Jailbreak Attack (arXiv 2404.01833)arXiv · Paper · published Apr 2, 2024 · accessed Oct 21, 2026
  5. Who's Harry Potter? Approximate Unlearning in LLMs (arXiv 2310.02238)arXiv · Paper · published Oct 3, 2023 · accessed Oct 21, 2026
  6. Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider (arXiv 2003.03423)arXiv · Paper · published Mar 6, 2020 · accessed Oct 21, 2026
  7. Singularity: Planet-Scale, Preemptive and Elastic Scheduling of AI Workloads (arXiv 2202.07848)arXiv · Paper · published Feb 16, 2022 · accessed Oct 21, 2026
  8. Confidential Consortium Framework: Secure Multiparty Applications with Confidentiality, Integrity, and High Availability (arXiv 2310.11559)arXiv · Paper · published Oct 17, 2023 · accessed Oct 21, 2026
  9. GRP-Obliteration: Unaligning LLMs With a Single Unlabeled Prompt (arXiv 2602.06258)arXiv · Paper · published Feb 5, 2026 · accessed Oct 21, 2026
  10. Lessons From Red Teaming 100 Generative AI Products (arXiv 2501.07238)arXiv · Paper · published Jan 13, 2025 · accessed Oct 21, 2026
  11. Securing AI Agents with Information-Flow Control (arXiv 2505.23643)arXiv · Paper · published May 29, 2025 · accessed Oct 21, 2026