# Nancy Wang

Nancy Wang is a technology investor focused on cybersecurity and AI. Wang has led investment in security companies and founded Advancing Women in Tech, a nonprofit that offers practical career education.

- Profile: https://b2binfluencer.marketing/creators/nancy-wang/
- LinkedIn: https://www.linkedin.com/in/wangnancy
- LinkedIn followers: 35,926
- Topics: artificial-intelligence, cybersecurity, startup, ai-agent-security
- Based on 15 dated posts, latest 2026-09-24

## Summary

Nancy Wang is most useful for practitioner conversations and advice on AI agent security and secure-by-design.

- Best for: A cybersecurity vendor launching AI agent security work needing a moderator, interviewer, or advisor.
- Not for: A consumer mobile app or lifestyle brand seeking high-volume paid social installs.
- Check first: Confirm current availability, vendor-sponsored work, conflicts or exclusivity, and audience composition.

### Takeaways

- Put security inside the AI generation loop, not bolt it on afterward.
- Treat every AI agent as untrusted and control what it can reach.
- AWIT partnered with Anthropic and Coursera for practical AI education.

## Main ideas

- **Put security inside the AI generation loop, not bolt it on afterward**: Wang argues that AI coding agents can produce insecure code even when the models understand security, because security is not the optimization target. Ambiguous specs and competing priorities let insecure code through, while loops that run faster than human review widen the gap between generation and review. Wang endorses embedding security constraints in the generation step: what the spec emphasizes shapes what you get, so security has to be part of the spec.
- **Treat every AI agent as untrusted and control what it can reach**: Wang's preferred posture is to assume agents can be compromised. That leads to concrete design choices: limit what an agent can access, monitor what it does, and put a policy enforcement layer at the boundary. They frame secure-by-default as a business philosophy, not a compliance checkbox, and describe agent identity as an unresolved industry question that companies are working around rather than waiting for.
- **Protect AI at runtime against concrete, named attacks**: Wang's argument about AI security as a market centers on speed and specificity: attackers are scaling faster than defenses, so defenses must evolve at the same rate. They name attack classes and when they occur, pointing to real-time detection and automated enforcement as the response.
- **Make overwhelming volumes of security data usable in real time**: Wang treats security data volume as one of the field's hardest practical problems, not a tooling detail. They value AI-native data pipelines because they let enterprise security teams work with telemetry as it arrives instead of drowning in it.
- **Convene practitioners and ask uncomfortable questions**: Wang advances the debate by assembling practitioners and pressing them on unresolved problems rather than restating that AI is changing everything. Their questions focus on what remains open: how an agent gets identity and authority, what differs between open and closed models, what recent exploits exposed, and where human control still has to stay.
- **Practical, workplace-ready education as a path into technology**: Alongside the security arguments, Wang emphasizes access: concrete, job-relevant skills can open technology careers, and the effort centers learners rather than prestige. They tie this to their experience growing up in rural Wisconsin, crediting small moments of access—a teacher, a program, scholarships—with opening a path that had not seemed available.

## Work to read

- [Black Hat panel on AI agent identity, authority, and control](https://www.linkedin.com/posts/wangnancy_im-thrilled-to-say-that-the-panel-im-hosting-activity-7488731004640100352-3M02)
- [1Password Zero-Shot Learning: treating agents as untrusted](https://www.linkedin.com/posts/wangnancy_new-1password-zero-shot-learning-is-out-activity-7476177417254903808-lBAn)
- [Corridor Series A: secure-by-design inside AI coding loops](https://www.linkedin.com/posts/wangnancy_these-26-year-old-founders-quit-dhs-and-built-activity-7440494002585505792-Mw29)
- [Operant AI and runtime AI security](https://www.linkedin.com/posts/wangnancy_aisecurity-runtimesecurity-cybersecurity-activity-7298542630806724608-JftM)
- [Observo AI's journey to SentinelOne](https://www.linkedin.com/posts/wangnancy_aiobservability-cybersecurity-rubrikmafia-activity-7370865272393678848-wqvZ)
- [Real-World AI for Everyone with Anthropic, Coursera, and AWIT](https://www.linkedin.com/posts/wangnancy_real-world-ai-for-everyone-available-now-activity-7399887896914092032-W6R4)

## Brand relationships

- **Advancing Women in Tech** (owned_business)
- **Felicis** (employment)
- **Corridor** (investment)
- **Observo AI** (investment)
- **Anthropic** (disclosed_partnership)
- **Coursera** (disclosed_partnership)
- **1Password** (disclosed_partnership)
- **Operant AI** (mention)
- **Black Hat** (mention)

## Partnerships

Nancy Wang operates in four partnership-shaped roles: hosting a practitioner panel at Black Hat, fronting a security interview series for 1Password, co-developing nonprofit AI courses through Advancing Women in Tech, and investing in early-stage security companies through Felicis. For a cybersecurity vendor weighing a speaker or advisor engagement on AI agent security, the closest precedents are the Black Hat panel, where they assembled CISOs, founders, and investors for unscripted practitioner discussion, and the recurring AI-agent-security theme in the investment and course work. Ask directly about pricing, turnaround, approval rights, deliverables, exclusivity, lead time, and any sponsorship policy or media kit.

## Alternatives

- Ronen Slavin: Choose Ronen Slavin when a campaign needs someone who can argue the remediation side of AI agent security in product terms, rather than host the conversation. The posts examine the same buyer problem from the defender's operational angle: AI writes code faster than humans review it; the 2026 Verizon DBIR shows vulnerability exploitation passing stolen credentials as the top initial access vector, median time-to-patch up 34% year over year, and supply chain involvement in 48% of breaches; and Forrester's first Agentic Development Security Landscape names the category being built to answer it. Slavin frames the architectural response as a context graph, an action layer with verifiable outcomes, and trust controls that hold when the agent is wrong, backed by a Mythos resource hub and guides on what to do before public model releases. If the vendor sells triage, exploitability confirmation, or closing risk at AI speed, an advisor already fluent in that stack can pressure-test the roadmap faster than a general moderator.
- Clint Gibler: Choose Clint Gibler when the campaign aims to reach working practitioners through an established reading habit rather than a single event. The two tl;dr sec issues reviewed curate research across AppSec, cloud security, supply chain, blue team, red team, and AI plus security, with concrete items such as why VMs will not contain cyber-capable agents, MCP configuration poisoning, a password-spraying campaign against AWS root accounts across 150-plus organizations, and autonomous AI agents breaking into online retailers for about $25 per target. For a vendor that wants its agent-security research alongside the links practitioners already scan each week, with a steady cadence rather than one panel, the format fits the same audience the AI agent security brief is aimed at.
- Motti Sorani: Choose Motti Sorani when the vendor needs a founder-level voice on why security work stops short of fixing anything, told through a personal origin story. One post traces learning assembly by reading computer viruses and building a virus generator as an intellectual challenge, before offensive security had that name; the other makes the remediation case directly: teams do not need more alerts, they need fixes that fit into development workflows, with a telecom-style last-mile analogy for the hardest part of delivery. The co-founder conversation format with Gil Regev gives the campaign a ready-made why-this-matters segment. For a vendor selling remediation automation or an application security workflow product, that combination of story and argument is more directly usable than a broader AI-security panel.

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