Corvus Insurance

As a member of the Executive team, Jason splits his responsibilities between the following functions:

 - Internal Security: Jason is responsible for defining and executing on a vision to secure the environment.
 - Product: Jason works closely with product, data science, and engineering teams to develop and execute a vision for a risk management platform. This includes passive security scanning technology and policyholder platform.

Chief Information Security Officer

NetDiligence®

Advisor

MOXFIVE

MOXFIVE is a Technical Advisory firm that helps organizations minimize the impact of current and future security incidents through strategic security initiatives and programs and incident management services. As the co-founder and COO of MOXFIVE, Jason defined the vision and built the foundation of MOXFIVE's Incident Management and Business Resilience services that help organizations bridge the gap between IT, Security, and Business leaders to drive better outcomes before, during, and after security incidents.

Co-Founder and Chief Operating Officer

Gigamon

After Gigamon acquired ICEBRG in July 2018, Jason led company-wide initiatives to assist with the integration of the two companies. These included expanding strategic partnerships, leading the Insight Channel GTM efforts, and building services powered by Gigamon technology that were delivered through Channel partners.

Senior Director of Channel Services

ICEBRG (acquired by Gigamon)

After building incident response teams, Jason become a key member of the ICEBRG leadership team. In his role at ICEBRG, Jason led the Strategic Partnership team to build, manage, and expand key partnerships with incident response firms who leveraged ICEBRG's network detection and response technology in investigations. Through these partnerships, Jason built a reliable sales funnel that supported the growth of the company through acquisition. Having completed the critical phase of integration, Jason left to co-found MOXFIVE.

Vice President of Strategic Partnerships

The Crypsis Group (acquired by Palo Alto Networks)

As a core member of the leadership team, Jason built the foundation of Crypsis's consulting services organization specializing in incident response. His areas of responsibility included recruiting and leading a team of 20+ consultants, driving thought leadership, and delivered cost effective incident response services for the cyber insurance industry,

Vice President of Professional Services

Mandiant

Jason led incident response investigations across numerous industries including the defense industrial base, financial industry, healthcare industry, Fortune 100 companies, and law enforcement. He became at expert conducting investigations into APT threat actors, financial crime, and hacktivists.

In addition to his day to day activities in responding to incidents, Jason also led training sessions for audiences in law enforcement, the federal government, and corporate security groups. He also helped organizations develop and improve incident response procedures and policies to assist in the detection and remediation of security incidents. 

Jason served as the regional incident response functional lead dedicated to continuous improvement of incident response processes and technologies to continue to support Mandiant's clients.

Manager

Rochester Institute of Technology

Jason was responsible for conducting vulnerability scans of the campus network to help secure the environment.

Information Security Engineer Associate

Senior Resident Advisor

Jason Rebholz is a cyber security leader passionate about dissecting complex problems and building programs from the ground up. His diverse experience in security start-ups contributes to his approach in leading and developing high-performing teams in fast-paced environments.

After spending a decade responding to sophisticated hacks, his professional mission evolved to solving cyber security at scale. In his current role as Chief Information Security Officer at Corvus Insurance, Jason blends his security expertise with cyber insurance to improve the security of thousands of companies.

** Comments, postings, and opinions expressed are solely my own **

Jason Rebholz

I'm thinking of using this as my new professional headshot. Thoughts?

I help companies secure AI | CISO, AI Advisor, Speaker, Mentor

Just another day in the Agentic AI office in 2026…

The best time to start securing AI is today. Don't let this happen to you.

If you’re looking to find a better way to secure AI beyond crossing your fingers that something bad doesn't happen, shoot me a DM. 

We’re onboarding early design partners to help shape the future of securing AI with us.

Security shouldn’t stop you from loading a cat meme.

That’s Perplexity’s philosophy when it comes to AI browsers…okay, I paraphrased it a bit. They just released a blog post on their new BrowseSafe defense-in-depth architecture. It includes:

▶️ 𝗧𝗿𝘂𝘀𝘁 𝗕𝗼𝘂𝗻𝗱𝗮𝗿𝘆 𝗘𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁: when the browser (or any connected tool) retrieves web content, a parallel detection pipeline is triggered.

▶️ 𝗛𝘆𝗯𝗿𝗶𝗱 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻: detection classifiers for “high-throughput scanning” look for prompt injection. If something is flagged, it is routed to “slower, reasoning-based frontier LLMs” for deeper analysis.

▶️ 𝗗𝗮𝘁𝗮 𝗙𝗹𝘆𝘄𝗵𝗲𝗲𝗹𝘀: when new threats reach the deeper analysis, they’re flagged to create new synthetic data to improve the layer-1 detection models. This creates a learning loop for continuous improvement.

Even with their improvements, they’re still finding some prompt injection attacks sneak by.

The most successful tactics?

1️⃣ 𝗦𝗵𝗶𝗳𝘁𝗶𝗻𝗴 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰𝘀: using multiple languages or indirect hypothetical framing has the largest impact on detection capabilities. Perplexity equated this to models overrelying on English triggers for detection.

2️⃣ 𝗛𝗶𝗱𝗲 𝗶𝗻 𝗣𝗹𝗮𝗶𝗻 𝗦𝗶𝗴𝗵𝘁: prompt injection hidden in the page, such as in HTML comments or data attributes, was easier to detect. They found that the hardest-to-detect attacks were embedded directly in text or elements that appeared on the page.

3️⃣ 𝗗𝗶𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗪𝗶𝗻: Detection accuracy dropped from 90% to 81% when adding in “benign command-like text.” The researchers noted that it suggests models are brittle in their detection capabilities.

Check out the full blog for a more detailed overview and where we need to go from here as an industry: https://lnkd.in/eT3c8br5

If you’re worried about securing AI applications or agents today, shoot me a DM. We’re building a better way at Evoke Security and are onboarding early design partners.

taplio