AI is Dangerously Outpacing Manual Security Workflows

Saket Modi
Co-founder and CEO

Q1. Cyber risk quantification is quickly becoming a mainstream board-level requirement. Where do you see organizations still struggling most in translating technical security data into financially meaningful risk decisions? How is Safe Security helping organizations address that challenge?

Organizations are not struggling because they lack data. They are struggling because they have too much of it. With Mythos and the broader rise of automated discovery, finding exposures is becoming table stakes. 87% of organizations have at least one known exploitable vulnerability in deployed services, affecting 40% of services (Datadog). The real challenge is no longer identifying issues. It is knowing which issues matter most.

Boards are not looking for another technical backlog. They need to understand which exposures could materially impact revenue, operations, customers, or regulatory standing. They need to know what to fix first, what risk reduction that investment will deliver, and how to make those decisions within budget. That is what will define cyber resilience going forward: the ability to prioritize the exposures that matter, reduce risk in a financially disciplined way, and prove ROI in terms the business understands.

This is where Safe Security helps. SAFE brings together Continuous Threat Exposure Management (CTEM), Third-Party Risk Management (TPRM), Artificial Intelligence Security Risk Posture (AI-SPM), and Cyber Risk Quantification (CRQ) into one platform that connects technical exposure to business impact.

With SAFE CTEM, organizations continuously identify, prioritize, and orchestrate remediation of the most critical exposures across their environment, linking tactical vulnerability management to systemic risk management. With SAFE TPRM, they understand which vendors create the greatest operational, financial, and compliance risk. With SAFE AI-SPM, they gain unified visibility into AI-related exposure across usage, configuration, outside-in posture, questionnaires, and contracts. And with SAFE CRQ, all of this context is translated into board-ready financial risk.

The result is an intelligent system that links strategy to execution, internal enterprise to external ecosystem risk, and data to measurable outcomes, giving leaders clear visibility, precise prioritization, effective remediation, and the ability to scale cyber risk management across the organization.

The result is a shift from “How many findings do we have?” to “Which risks matter most, what will it cost to reduce them, and how much risk reduction will we achieve?” That is what the conversation boards need, and what SAFE enables.

Q2. As Mythos makes exposure discovery more continuous and scalable, what will separate organizations that are truly “Mythos-ready” from those that simply generate more vulnerability data? How is SAFE helping security teams prioritize and reduce the exposures that matter most?

Today, most organizations struggle to manage even a few hundred “critical” vulnerabilities across their environment. With Mythos and AI-driven discovery, that number could quickly go up by 3 to 5x. But the team responsible for managing that risk does not become 3 to 5x larger. Budgets do not expand at the same pace. And not every exposure will have a patch, require a patch, or deserve immediate action.

That is why CTEM must evolve from a visibility program into an autonomous risk reduction program.

A truly Mythos-ready enterprise needs to continuously answer: Is this exposure real? Is it exploitable in our environment? Does it affect a critical asset or business process? What should we do next? And how do we prove that risk went down?

SAFE makes enterprises Mythos-ready by creating one trusted view of exposure risk. It brings together fragmented exposure data across tools and normalizes it into a unified exposure knowledge graph. SAFE prioritizes exposures using risk context such as exploitability, threat intelligence, asset criticality, business impact, access, and compensating controls. This helps teams focus on what matters most instead of chasing every technical finding. SAFE also helps organizations understand what is truly reachable and exploitable, whether existing controls reduce the risk, and what its financial impact could be.

But the real differentiator is SAFE’s Agentic Workflow Engine. Powered by 100+ AI Agents, SAFE does the heavy lifting across the CTEM lifecycle. It automates exposure prioritization, business impact analysis, ownership routing, ticket creation, exception management, remediation guidance, follow-ups, and executive reporting. Instead of analysts spending cycles deduplicating findings, chasing owners, reconciling tools, and preparing status reports, SAFE enables them to focus on exposure management and risk reduction.

Finally, SAFE gives leaders board-ready visibility into whether CTEM is actually reducing risk. Through CRQ-powered reporting and SafeX conversational AI, teams can instantly answer questions about high-risk exposures, crown jewels, owners, SLAs, MTTR, backlog trends, and financial risk reduction. In the Mythos and AI era, CTEM readiness is not about finding more. It is about deciding better, acting faster, and proving measurable exposure reduction. SAFE enables that shift.

Q3. What are Safe Security's plans at Black Hat USA 2026? What innovations do you plan on showcasing at the event?

At Black Hat USA 2026, SAFE’s focus is on Autonomous Cyber Risk Management because the risk environment has changed faster than the operating model used to manage it.

AI adoption is accelerating, and at the same time, AI is expanding the attack surface. Even the White House is elevating AI security as a national priority, underscoring that AI innovation and AI risk now have to be managed together. The challenge for enterprises is no longer whether they can find risk. Mythos and AI-driven discovery will make exposure visibility more continuous and more scalable. The challenge is whether teams can understand, prioritize, and reduce the right risk fast enough, without adding endless manual effort.

That is what SAFE is showcasing at Black Hat: an autonomous operating model across SAFE CTEM, SAFE TPRM, SAFE AI-SPM, and SAFE CRQ, powered by Agentic Workflows and 100+ AI Agents. SAFE represents a fundamentally new way to manage cyber risk, purpose-built as a cyber risk decision engine for modern enterprises. SAFE unifies Cyber Risk Management (CRQ), Continuous Threat Exposure Management (CTEM), and Third-Party Risk Management (TPRM) into a single platform.

With SAFE CTEM, organizations can move from vulnerability volume to validated, business-aligned exposure reduction: unifying exposure data, identifying what is truly exploitable, and orchestrating remediation.

With SAFE AI-SPM, we are helping enterprises govern AI adoption without slowing innovation. SAFE gives visibility across AI usage, configurations, outside-in exposure, questionnaires, and contracts, so organizations can understand where AI risk is emerging and act before it becomes systemic.

With SAFE TPRM, we are addressing the scale problem in third-party risk. Vendor risk can no longer depend on static questionnaires and manual follow-ups. SAFE helps teams continuously assess vendor exposure, identify the riskiest third parties, and connect vendor risk to potential financial loss.

And with SAFE CRQ, all of this becomes financially meaningful: where risk is concentrated, what actions reduce it, and how cyber investments improve resilience.

SAFE is on a mission to build Cybersecurity Superintelligence that becomes a reasoning layer and the central nervous system of enterprise cybersecurity.

Eric Doerr
Chief Product Officer

Q1. Tenable recently launched Hexa AI? What exactly is it? How exactly does it help organizations keep pace with the rapid vulnerability discovery enabled by frontier models like Anthropic’s Mythos?

Human execution capacity is the scarcest resource in the enterprise. As frontier models compress vulnerability discovery from months to minutes, manual security workflows are being dangerously outpaced. Manual triage becomes a losing battle when attackers can move at machine speed.

Tenable Hexa AI is the agentic engine inside the Tenable One Exposure Management Platform that translates exposure data into prioritized, business-aligned intelligence and coordinated remediation action. Customers are using Tenable Hexa AI to automate assessment configurations, orchestrate complex cross-domain workflows and automatically patch software, change configurations or modify code the moment an exposure is validated. By connecting directly to your technology stack, mapping complex relationships across the attack surface and executing workflows automatically, Tenable Hexa AI enables organizations to neutralize risk before it can be exploited. It ensures your best people are no longer buried in manual triage, giving them the leverage to harden your defenses and close the cybersecurity gaps that put your business at risk.

Q2. What will effective cyber risk prioritization actually look like in practice over the next few years as attack surfaces become more dynamic and identity-driven? How will it be different from risk prioritization as practiced today?

Calculating risk across dynamic environments is like finding a needle in a haystack. Historically, the security industry has operated on a fundamentally flawed assumption: that we can only fix a tiny fraction of our vulnerabilities. We spend massive amounts of energy filtering millions of alerts down to a top ten list, when resources should be hyper-focused on resolving issues based on the likelihood of attack. In fact, according to Tenable Research, only 3% of exposures pose a true risk to businesses.

Tenable One and Tenable Hexa AI are equipped with advanced prioritization capabilities. Fueled by the Tenable Exposure Data Fabric, the industry's largest repository of contextualized exposure data, Tenable normalizes, deduplicates and processes exposure data from native telemetry, technology partners and custom data sources to deliver a clear picture of all assets and exposures in one place. This view of risk, threat intelligence and deep understanding of the complex relationships between assets, environments and exposures powers Tenable One to accurately and effectively prioritize action based on real business impact. Instead of security teams playing a high-stakes game of whack-a-mole, they confidently reduce risk wherever it lives.

Over the next few years, we’ll see organizations becoming increasingly comfortable implementing AI into remediation workflows. Hexa ensures the customer is in control and can move at the “speed of trust” - fully automated, human in the loop, or any hybrid.

Q3. Tenable packed in a lot at Black Hat USA 2025. What are your company's plans at the event this year? What new announcements, experiences or key themes will you be showcasing to attendees at Black Hat USA 2026?

The explosion of AI-driven tools has created a crisis of complexity, paralyzing security operations with fragmented tools and workflows. This siloed data makes coordinating a unified defense nearly impossible. At Black Hat USA 2026, Tenable is demonstrating the next evolution of exposure management to solve this exact problem.

We are highlighting Tenable Hexa AI as the catalyst for autonomous operations. As frontier models unearth thousands more vulnerabilities than what we’ve become accustomed to, organizations can outpace those threats at scale with Tenable Hexa AI. We are bridging the critical gap between vulnerability discovery and remediation. Attendees can stop by the Tenable booth (#2639) to see Tenable Hexa AI in action.

Tenable's Black Hat presence will include a major booth with live demos and access to our product experts, so customers, partners and prospects can learn more about our industry-leading AI-powered exposure management platform. By empowering hybrid teams of humans and AI agents to execute seamlessly together, we will show attendees exactly how to cut through the noise and take complete control of a rapidly expanding attack surface.

Alon Schindel
VP of AI & Threat Research at Wiz

Q1. What are the most concerning new attack patterns and inherited risks you’re seeing in cloud environments today? How are sophisticated attackers exploiting the speed and autonomy of AI systems?

Attackers are not using fundamentally new techniques. They’re using familiar ones faster and at greater scale. AI is reducing the cost of discovery, accelerating exploit development, and compressing the time between vulnerability discovery and exploitation. Findings from our latest State of AI in the Cloud Report highlight several trends shaping cloud environments today.

First, autonomous agents are expanding the attack surface in ways many organizations have not fully mapped. At least 57% of organizations have deployed self-hosted AI agents, and MCP servers now appear in 80% of environments. These orchestration layers introduce significant control plane risk. If an agent is overprivileged or exposed to the internet without proper guardrails, attackers can hijack that access to move laterally across sensitive systems and data stores.

Second, AI-assisted development is creating systemic inherited risk. More than 80% of organizations use AI IDE extensions, and 71% have at least one AI coding assistant in use. As AI-generated code and configurations are reused across environments, small mistakes can spread quickly. Roughly one in five organizations using AI-powered coding platforms had applications affected by issues tied to shared generation patterns.

Third, attackers are operationalizing LLMs within malware and intrusion workflows. Some malware families now use LLMs to dynamically generate commands and adapt execution logic at runtime, reducing reliance on static payloads. Attackers have also abused AI-enabled OAuth integrations to move laterally across SaaS environments through trusted automation paths.

Frontier models have also demonstrated the ability to autonomously identify vulnerabilities and generate working exploits, further shrinking the exploitation window. Security teams need to operate with that assumption today.

Q2. What’s the most critical strategic shift security leaders need to make in 2026 to secure workloads in the cloud? Are there specific capabilities, processes, or metrics that separate the leaders from those still playing catch-up, in this regard?

The most important shift is adapting to continuous, AI-driven discovery and exploitation.

The window between exposure and exploitation is shrinking. Security teams need to continuously reduce the time between identification, validation, and remediation across code, infrastructure, and runtime. Fast patching becomes a core part of reducing risk.

Organizations that struggle to keep pace often face the same challenges: incomplete visibility and slow remediation. They can identify exposed assets, but lack the context to determine what’s actually exploitable. They surface issues, but ownership and remediation paths remain fragmented across teams.

The risk is measurable. Our data shows that 30% of cloud environments have at least one high-impact machine running software that is exposed externally. Even if that software is not exploitable today, the pace of AI-driven vulnerability discovery means it is likely only a matter of time before it becomes exploitable.

What separates mature programs is operational consistency around speed of action and breadth of visibility. Continuous asset discovery with clear ownership, validation of exploitable attack paths, defined remediation workflows, and measurable reductions in attack surface over time all help teams move to a place of AI threat readiness. The goal is not simply to move faster. It’s to build an operating model where continuous visibility, prioritization, and remediation become part of how the organization operates by default.

Q3. How is Wiz planning to engage with customers, partners, and the broader security community at Black Hat USA 2026? Are there any key initiatives or announcements you plan on making at the event?

Security in the AI era requires a new operating model. For the first time, defenders can move as fast as attackers, without sacrificing precision or slowing innovation.

At Black Hat, we’ll share more about how organizations can build AI threat readiness into their operating model. We’ll also share more learnings from our Wiz Red Agent, an external AI attacker that has been continuously learning and refining its adversarial capabilities across an ever-expanding dataset of over 150K production web applications and APIs scanned weekly.

Wiz researchers will also lead several briefings at Black Hat, covering topics ranging from the takeover of a flagship cloud service to the investigation of a sophisticated multi-cloud cryptocurrency theft. These sessions reflect our ongoing commitment to advancing cloud and AI security research and sharing those findings with the broader security community.

Shon Myatt
Chief Technology Officer

Q1. How do you see the concept of external attack surface management evolving over the next few years? What capabilities do you think will become table stakes in the space, and where do you see opportunities for meaningful differentiation emerging?

The lines between attack surface management, digital risk protection, and threat intelligence are blurring, and the space is consolidating into a single offering built around exposure. Gartner already treats standalone EASM as a feature of that, not a market of its own. Calling yourself an attack surface company in a few years will sound like calling yourself a firewall company today. Consolidation won't reward the biggest scanner. It rewards whoever can run the whole loop, from the first signal to the action that shuts it down, across both the digital and the physical.

The entry fee climbs past the data access fee. Anyone can scan what's visible, and access to the closed side, the underground forums and broker markets, is becoming table stakes too. What separates the field is what you can do with what you find, and whether you understand what you're looking at. Taking action means takedowns and disruptions that stay within legal and operational boundaries, not hacking back. Understanding means real context on the customer: their brand, their executives, their assets, and how adversaries actually come at them. Exposure without that context is just a longer list of things to worry about.

So differentiation isn't any single piece, because the pieces can be bought. It's the combination. Being inside the adversarial networks where credentials and footholds change hands, and having both the customer's trust to act on their behalf and the partner network that makes a takedown happen, the registrars, hosts, and platforms that pick up the phone. Under it all sits a decade spent protecting brands, executives, and the people around them, digital and physical, as one problem. A well-funded entrant can buy the data and train a sharper model. It can't shortcut relationships. Not what we can see, but who we're connected to, and the reach to take action on a threat once we find it.

Q2. AI is dramatically lowering the barrier for creating convincing fake identities, fraudulent campaigns, and synthetic personas at scale. What does that shift mean for how organizations establish digital trust and verify authenticity online going forward?

Impersonation used to be hard to pull off at scale, which helped protect against it. What took a team and weeks, whether faking an executive or a whole brand, now takes one operator and an afternoon:a cloned voice, a spoofed domain, a synthetic face built from whatever the target has posted. These attacks don't exploit a flaw in your systems. They exploit trust, and they arrive through the channels you trust most: a video call, a note from an executive, a familiar logo, the exact places no one stops to verify. Signature and pattern detection breaks when every fake looks different, and the volume keeps climbing, and pure detection is a fight the defender loses as the models keep improving.

Provenance is the industry's answer, and it has real limits. Content Credentials can prove how a file was made, but they get stripped the moment content crosses the social and messaging platforms where impersonation actually lives, most of which still don't carry the standard. Provenance helps the honest publisher, but does little against an adversary who never signs anything. So trust has to be rebuilt in how organizations operate, through multi-channel confirmation before anyone acts on a sensitive request, and validation for any content that shows an executive saying something consequential. It also means shrinking the raw material itself, because a leader with hours of public audio and video and personal details sitting across data brokers is the easiest person in the company to clone and find.

The question itself has to change. Not "is this real," but "who's behind it, and what's it part of." Spotting a fake is table stakes; tracing it back to the actor, then taking it down where you can and alerting where you can't, is the work that matters. Let machines flag the anomalies at attacker speed, because the real edge is the access itself, the standing inside closed networks that lets our agents attribute a campaign and act on it. A synthetic version of an executive starts as a trust problem and becomes a safety problem the moment it's used to move money or put a stranger outside someone's home, which, for the target, were never separate problems to begin with.

Q3. What are your company's priorities at Black Hat USA 2026? What do you want attendees to take away from your organization's participation at the event?

The conversation we want to have at Black Hat is one the industry is overdue for: the line between digital and physical risk has disappeared, and most security programs aren't built for it.

Security teams are still treating Cyber Threat Intelligence and Physical Security Intelligence as separate disciplines, with separate teams, separate tools, and separate mandates. That boundary doesn't exist for attackers. A layoff announcement becomes a grievance thread becomes a doxxing campaign. A Zillow listing becomes a home visit. A protest gets organized online, geo-tagged to an executive's office, and nobody in the SOC is watching. Each of those scenarios crossed a boundary most traditional programs won't catch.

We want attendees to leave with a clear picture of how threats move from a digital signal to a physical outcome, where most security programs lose the thread, and what a mature, converged operating model looks like. Executive protection that only monitors a single identity misses the family, the residence, the travel, and everything the executive discloses about themselves online.

And we'll be pulling back the curtain on HNTR, our newly rebuilt platform, and HNTR Executive Protection, the application built to run exactly this kind of convergence.