As Black Hat USA 2026 prepares to descend on Las Vegas, the security conversation has shifted decisively from “shadow AI” to a thornier problem: the AI agents enterprises have already built, sanctioned, and pushed into production. Straiker, which bills itself as The Agentic Security Company, is arriving with a pointed message for CISOs — the agents already running in your environment, with the access you already granted them, are the real exposure, and most organizations have no visibility into what those agents are doing at runtime. Ahead of the show, VMblog caught up with Amy Heng, head of marketing at Straiker, to talk about the company’s approach to discovering, red-teaming, and defending AI agents across the full agentic surface, from MCP servers to autonomous workflows.
In this Q&A, Heng walks through what attendees can expect at Straiker’s booth this year, including a live demonstration of Ascend AI and a hands-on “agentic kill switch,” along with news the company is teasing around a recently closed Series A and its first agentic threat research report. She also digs into the threat vectors Straiker sees defining the second half of 2026 — prompt injection, tool abuse, and agent hijacking chief among them — and offers a contrarian take on identity security, arguing the deeper danger lies in indirect prompt injection quietly steering agents that already hold legitimate access. It’s a candid look at where one vendor believes the industry’s blind spots really are, just as agentic AI moves from experimental to enterprise-critical.
++
VMblog: For readers who may not be familiar, give us the elevator pitch — who you are, what you do, and what genuinely sets you apart in today’s crowded cybersecurity market.
Amy Heng: Straiker is The Agentic Security Company. We help security teams discover every AI agent and tool connection in their environment, red-team those agents continuously with real adversarial attacks, and defend them at runtime, so the people shipping agents can move fast without shipping risk. What sets us apart: while most tools watch the model, we secure everything the agent actually does above it.
VMblog: Black Hat attendees are a discerning crowd. What booth experiences, live demos, or hands-on activities are you bringing to Las Vegas that will cut through the noise and leave visitors with something they can’t stop thinking about?
Heng: We’re excited to show off Ascend AI live and hand you an agentic kill switch, because the more useful and complex an agent becomes, the more dangerous it is to an enterprise, and control belongs back in human hands. Come watch us red-team an autonomous agent in real time, then flip the switch and take control back.
VMblog: What’s your Black Hat origin story? Longtime exhibitor or fresh face on the floor? What keeps your company coming back — or what made 2026 the year to finally plant your flag here?
Heng: Last year, we were at the Start Up Village. This year? We’ve landed ourselves a 20×20 booth, and right next to our biggest competitor. You should come by and see why we keep beating them.
VMblog: If a CISO walks away from your booth remembering exactly one thing about your company, what do you want that to be?
Heng: The smarter your agents get, the more dangerous they get, and right now most of them are running with the keys and no supervision. Straiker gives you the map, the red team, and the kill switch, so humans stay in charge.
VMblog: The threat landscape heading into Black Hat 2026 looks very different from even 18 months ago. Which specific threat vectors — whether that’s agentic AI attacks, identity-based intrusions, critical infrastructure targeting, or something else — is your solution most directly built to address?
Heng: We’re built for agentic AI attacks: prompt injection, tool and function abuse, agent hijacking, data exfiltration through autonomous workflows, and the identity sprawl that comes when agents act on their own. These are the vectors that didn’t meaningfully exist 18 months ago and now sit in production.
VMblog: Agentic AI is reshaping both offense and defense. How is your company building security for — and with — autonomous AI systems, and what risks are you most concerned enterprises are underestimating right now?
Heng: Straiker is agentic by design and runs as a closed loop: we red-team your agents with a unique adversarial dataset built through our work with frontier AI labs, then feed everything we learn straight into runtime visibility, governance, and enforcement across your agents, MCP servers, and Skills. We began as runtime defense, opened our own offensive testing to customers, and now cover the full agentic surface. What enterprises underestimate is that a manipulated agent doesn’t just answer wrong; it acts, with all the access you already gave it.
VMblog: “AI-native security” is quickly becoming the new “next-gen.” What does that phrase actually mean at your company, and how do you demonstrate real differentiation beyond the marketing language?
Heng: For us, it means the product secures AI using AI, not a legacy scanner with an AI label. We prove it live at the booth: continuous adversarial testing at machine speed and runtime enforcement that a rules-based tool can’t match, refined constantly from what our customers see in the wild.
VMblog: Are you unveiling any major product announcements, partnerships, or research findings at Black Hat 2026? Can you tease it or give us a preview?
Heng: The last 60 days have been big for us: we closed our Series A with top-tier investors such as Marathon, Citi Ventures, Workday Ventures, and Illuminate Financial, including continued support from BCV and Lightspeed and launched our first threat research report on agentic attacks. Come by the booth for a preview of both, plus new Ascend AI capabilities, and catch us at Ai4 as well.
VMblog: Identity has become the new perimeter — and attackers know it. How has your approach to identity security, authentication, or access management evolved, and what are organizations still getting dangerously wrong?
Heng: Here’s our contrarian take: identity isn’t the core problem, because an agent already inherits the permissions of whoever commands it. The real gap is indirect prompt injection, a poisoned file, a calendar invite with hidden instructions, a README with a buried agenda, quietly steering an agent that’s operating with fully legitimate access.
VMblog: What’s the most significant cybersecurity blind spot you’re seeing across your customer base right now, and how does your technology address it?
Heng: Everyone’s hunting shadow AI while the real exposure is the agents you sanctioned and shipped yourself. Once they’re in production, they don’t trip DLP, don’t show on a usage dashboard, and don’t ask before they act, and that’s exactly the layer Straiker discovers, tests, and defends.
VMblog: Security teams are being asked to do more with tighter budgets and leaner headcounts. How does your solution help security leaders justify ROI and actually reduce operational burden rather than add to it?
Heng: Straiker automates the red-teaming and runtime defense that teams can’t staff manually, so lean teams cover far more attack surface without adding headcount. The ROI story is simple: continuous coverage of a fast-growing risk at a fraction of the cost of catching it after an incident.
VMblog: What should be sitting at the very top of every security leader’s priority list in the second half of 2026?
Heng: Get agentic security right now, before it’s everywhere, because anyone can build an app or bolt AI onto an existing one, often with a coding agent doing the work. That explosion of AI-enabled apps and the coding agents building them is the H2 priority: red-team them the way attackers actually will, and enforce guardrails at runtime before a hidden instruction turns that access into an action.
VMblog: The conversation around software supply chain security has matured significantly — but has enterprise practice kept pace? What’s the current state, and where are the gaps that still keep you up at night?
Heng: Practice has improved on traditional dependencies, but the AI supply chain, models, agents, and MCP integrations is the new soft spot. Those components pull in behavior and data at runtime, and few teams have any assurance over what they’re actually doing.
VMblog: Looking toward 2027 and beyond, what emerging threat or technology inflection point do you think the industry is still not taking seriously enough?
Heng: Billions of AI agents are about to join the enterprise workforce. AI training 1.0 was reading the world’s corpus; the next wave learns from real work and then executes it, so agents will act with more authority every quarter. Security has to move in lockstep with that, or it gets left behind the moment agents start doing the work.
VMblog: Does your team have any speaking sessions, sponsored research presentations, or Briefings appearances at Black Hat 2026 that attendees should put on their schedule?
Heng: Come by our booth!
VMblog: Beyond the product pitch, what’s one piece of hard-won, actionable security wisdom you’d hand to every practitioner who stops by your booth?
Heng: Remember that with agents, useful and dangerous grow on the same curve: the more an agent can do, the more it can be turned against you. Treat every new capability as a new attack surface, and test it before you trust it.
VMblog: Are you hosting any exclusive networking events, hospitality suites, or invite-only dinners during Black Hat week? How should interested attendees get connected?
Heng: We will provide the party-hards an IV drip and oxygen bar (plus breakfast) on Wednesday and Thursday during Black Hat: Wednesday: https://luma.com/4m7a360q; Thursday: https://luma.com/ubb77v65
VMblog: What’s the most creative or unexpected giveaway your booth is bringing this year?
Heng: Custom ASMR keyboards and agentic “kill switch” fidget toys, because securing autonomous systems should feel as satisfying as flipping the switch.
VMblog: For the first-timer navigating Black Hat for the first time — what’s your best advice for getting maximum value out of the week without burning out by Wednesday?
Heng: Pick three things you actually want to learn, wear comfortable shoes, and stop by Straiker’s Ascend IV and Runtime Recovery on Wednesday or Thursday morning to reset after the night before.
VMblog: When the industry gathers again at Black Hat 2027, what do you think will be the defining cybersecurity conversation that dominated the year?
Heng: AI agents. Who’s securing them, who’s exploiting them, and whether the industry built the guardrails in time.
##
