# Deepen Desai

Deepen Desai works on cybersecurity at Zscaler and shares practical perspectives on AI security, threat research, phishing, and security operations.

- Profile: https://b2binfluencer.marketing/creators/deepen-desai/
- LinkedIn: https://www.linkedin.com/in/deepen-desai-48487b2
- LinkedIn followers: 9,918
- Topics: artificial-intelligence, cybersecurity, agentic-ai, ai-security, zero-trust
- Based on 10 dated posts, latest 2026-10-03

## Summary

Deepen Desai is most useful for translating AI security, phishing, and security operations research into defensive priorities.

- Best for: A security vendor seeking a practitioner voice for co-authored threat research or a webinar.
- Not for: A vendor wanting an independent, vendor-neutral paid reviewer to rank competing endpoint or SecOps platforms.
- Check first: Confirm whether Zscaler's security organization can approve the partnership and if an independent booking route exists.

### Takeaways

- AI models can build complex exploit paths, so defense must assume machine-speed attacks.
- Phishing volume is falling but methods improve; priorities include segmentation, deception, phishing-resistant MFA, awareness.
- Security operations should close the loop rather than add more tools.

## Main ideas

- **AI models can now build attacks, so defense has to assume machine-speed exploit paths**: Deepen Desai argues that frontier AI has moved beyond scanning for known vulnerabilities: with multi-step reasoning, a model can autonomously construct a complex exploit path. Two observations support that claim: architectural context significantly improves a model's accuracy when exploiting a target, but that context should ground it in its own environment, not extend beyond it. The conclusion calls for a combination of controls, not a single product: AI-powered scanning across the software development lifecycle and CI/CD pipeline, Zero Trust architecture, AI-powered exposure management, and deception technology.
- **Phishing is falling in volume but improving in method, and the countermeasures are a short, named list**: Deepen Desai reads the annual Zscaler ThreatLabz phishing and initial access data as a story of quality over quantity. Volume dropped about 20% year over year, but campaigns increasingly combine email with phone calls, voice, or text-based social engineering. Attackers are moving beyond credential theft toward session tokens and rapid post-compromise action, while agents are being phished by humans and other agents. The report sets out three concrete priorities in response to these changing methods.
- **Security operations needs to be rebuilt around closing the loop, not adding more tools**: Deepen Desai argues that the traditional SOC model was already strained and is now breaking down: fragmented tools, incomplete visibility, high data-ingestion costs, and too many manual steps leave responders too slow against automated attacks. The proposed fix closes the gap between signal and action: unified exposure and threat management in one platform; operational access to inline telemetry across network, identity, endpoint, cloud, AI, and deception; specialized AI agents shaped by frontline SOC, managed detection and response, and threat-hunting experience; and closed-loop remediation through the Zero Trust Exchange, turning detection into real-time action.
- **Practical preparation should come from tested research partnerships, presented with named colleagues**: Deepen Desai turns threat research into practitioner guidance by naming the shift, citing research programs that granted early access, and promising practical preparation rather than theory. Webinars and shared research bring the work to practitioners alongside named colleagues.

## Work to read

- [Zscaler ThreatLabz 2026 Phishing and Initial Access Report: what Deepen Desai says defenders should take from it](https://www.linkedin.com/posts/deepen-desai-48487b2_cybersecurity-ai-securityleadership-activity-7470958537180028928-Fyoy)
- [“When the scanner starts thinking”: testing frontier AI models against real attack and defense scenarios](https://www.linkedin.com/posts/deepen-desai-48487b2_when-the-scanner-starts-thinking-learnings-activity-7463682090065313792-1Bf-)
- [Launching Zscaler Agentic Security Operations: why the SOC model is under pressure](https://www.linkedin.com/posts/deepen-desai-48487b2_ai-actfaststaysecure-cio-activity-7504265257604009985-Vvx3)
- [Webinar: practical guidance for security teams facing machine-speed attacks](https://www.linkedin.com/posts/deepen-desai-48487b2_a-practitioners-guide-to-secure-against-activity-7462973598413193216-aCky)

## Brand relationships

- **Zscaler** (employment)
- **Anthropic** (mention)
- **OpenAI** (mention)

## Partnerships

Deepen Desai’s posts show Zscaler work, not personal partnership or sponsorship formats: a company webinar on AI and vulnerability management with Zscaler colleagues Dhawal Sharma, Zoltan Kovacs, and Amir Moin; a Zscaler product launch featuring speakers from Anthropic, HDFC Bank, MAIRE, and Forrester; Zscaler’s security team gaining early access to frontier models through Anthropic’s Project Glasswing and OpenAI’s Trusted Access for Cyber program; and the company’s phishing report and threat-lab blog. These are employer-level relationships announced by an employee. For vendors planning a webinar or co-authored threat-research project, a pitch to Desai is effectively a pitch to Zscaler’s security organization, where public-facing thought leadership is produced under the company brand and partners are announced by the company. Ask for confirmation of the appropriate contact, sponsorship terms, rates, and deliverables; the posts provide no sponsorship policy, rate card, media kit, contact route, or disclosed paid collaboration.

## Alternatives

- Jay Chaudhry: Choose Jay Chaudhry for a top-level executive keynote on securing enterprise AI at scale, not a narrow research walkthrough. Jay Chaudhry's posts show a strategy-first argument: AI security cannot be paced even if frontier-model development is debated, traditional firewall-based security is unsafe against AI agents, and Zero Trust should hide attack surfaces and stop lateral movement. The Gartner summit post adds named customer stories from Dow and Viatris and frames the coming network as millions of autonomous AI agents. That is a stronger choice for a CIO/CISO main-stage session that needs executive-level stakes, named enterprise examples, and a call to reduce organizational inertia.
- Ronen Slavin: Choose Ronen Slavin for a campaign focused on application security, software supply chain, or remediation at machine speed. Slavin's posts treat the problem as a bottleneck shift: AI has commoditized vulnerability discovery, so the harder job is triaging, confirming exploitability, and closing risk as fast as AI finds it. One post uses Verizon DBIR findings, including vulnerability exploitation passing stolen credentials, time-to-patch up 34%, and supply chain involvement in 48% of breaches, alongside Forrester's first Agentic Development Security Landscape to define the category. The other covers Anthropic's Fable5 and Claude Mythos 5, Project Glasswing, OpenAI's Daybreak, and Microsoft's MDASH, all landing in the same remediation queue. That is a sharper fit than Desai for a vendor whose product or data sits in AppSec, secure SDLC, or vulnerability remediation.
- Sam Curry: Choose Sam Curry for leadership, workforce, and risk management in an agentic AI program rather than technical threat research. The posts ask how to lead non-human team members, maintain accountability when decisions are made by a network of agents, and keep predictability in non-deterministic systems. One post points readers to an article on those leadership challenges; the other is a conversational interview with Richard Stiennon that ranges from feature usage to information use and the societal and cyber impact of agentic AI. That is a better choice for a webinar or co-authored piece aimed at CIOs and CISOs who must reorganize teams and governance before technology outpaces management.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/