Alex: Welcome to Cleartext for Wednesday, September 2nd, 2026. I'm Alex Chen.

Jordan: And I'm Jordan Reeves. Alex, I want to start with something that should have every CISO checking their edge infrastructure this morning. SonicWall disclosed two zero-days in their SMA 1000 series, one of them a perfect ten CVSS pre-auth SSRF, and they chain together for remote code execution. Both actively exploited in the wild. If you're running SMA 1000 appliances, stop what you're doing and patch. We'll come back to that, but I wanted it said up front.

Alex: Absolutely, and we've got a packed show today. We're going to dig into China's Fire Ant campaign turning Cisco routers into persistent attack platforms, OpenAI releasing a model they themselves classify as having critical offensive cyber capabilities, a massive identity verification breach that could undermine KYC processes across entire industries, CISA pulling back free assessment services, and the growing trend of ransomware groups recruiting insiders. A lot to cover, so let's get into it.

Jordan: Let's start with Fire Ant because this one is architecturally significant. This is a China-nexus campaign that compromised Cisco routers, not as endpoints, but as pivot platforms, persistent infrastructure for launching further intrusions. The researchers were very deliberate in their language. They said this didn't just compromise systems, it compromised the trust layer those systems depend on. And that framing matters.

Alex: It matters enormously because think about how most enterprises treat their routing infrastructure. These devices sit in the most trusted positions in the network. They're making forwarding decisions, they're handling BGP sessions, they're the backbone. And yet they often get less security scrutiny than a user's laptop. How many organizations have integrity monitoring on their router firmware? How many are doing regular validation that the IOS image running is the one they deployed?

Jordan: Very few. And that's exactly what makes this class of attack so effective. You're not tripping EDR. You're not touching an endpoint. You're living in the infrastructure layer that everything else trusts implicitly. The attribution to a China-nexus actor is consistent with what we've been tracking for years, going back to Volt Typhoon, Flax Typhoon, and now Fire Ant. The pattern is prepositioning in network infrastructure for persistence and future access.

Alex: The board-level conversation here is about trust assumptions. If your routing infrastructure is compromised, your segmentation is illusory, your logging may be incomplete, and your incident response playbooks probably don't account for it. CISOs need to be asking their network teams hard questions about edge device integrity verification, firmware signing validation, and whether they have any ability to detect unauthorized modifications to router configurations or images.

Jordan: And I'd add, if you're running end-of-life Cisco gear that's no longer receiving security updates, this is your wake-up call to accelerate that refresh. You cannot defend what you cannot patch.

Alex: Let's pivot to what I think is the most consequential macro story of the week, possibly the quarter. OpenAI is about to release its Astra model, and they've classified it internally as having critical offensive cyber capabilities. That's their own assessment, not a researcher's, not a journalist's framing. OpenAI is telling select partners, get your defenses ready before we release this publicly.

Jordan: I've been watching the AI capabilities curve for years, and this is the inflection point a lot of us predicted. When the model developer itself acknowledges critical offensive capability and gives partners a head start on defense, that tells you the asymmetry is about to shift. The question isn't whether threat actors will use AI-assisted attack tooling. They already are. The question is what happens when a model this capable becomes broadly accessible.

Alex: And for CISOs, this changes the threat modeling conversation fundamentally. Your red team exercises, your assumed attacker capability profiles, your detection engineering, all of that needs to be recalibrated. If a model can autonomously identify vulnerabilities, craft exploits, and iterate on evasion techniques, the velocity of attacks changes, not just the sophistication.

Jordan: Which makes the HiddenLayer funding story worth a quick mention here. They just raised a hundred million Series B with Microsoft M12, Morgan Stanley, and Booz Allen on the cap table. That investor list tells you everything. When defense contractors and the hyperscaler that runs Azure are backing AI model security, this isn't a niche play anymore. It's becoming a board-level procurement category. CISOs who don't have an AI security strategy are going to be answering uncomfortable questions very soon.

Alex: Agreed. And the timing is not coincidental. The market is responding to exactly the kind of capability escalation that Astra represents. Let's move to the breach stories because they're connected in ways that matter strategically.

Jordan: The big one is this FBI investigation into a dark web service selling over 153 million U.S. and Canadian driver's license scans. Krebs traced it back to what appears to be a Louisiana-based identity verification company. This isn't a breach of a government database. This is a breach of the private sector intermediary that governments and enterprises trust to verify identity.

Alex: And that's what makes it structurally damaging. Think about how many processes depend on driver's license verification. KYC for financial services, patient identity in healthcare, employee onboarding, age verification. If the images in this dataset are high-fidelity, and early reporting suggests they are, you've just undermined a foundational identity proofing mechanism across multiple industries.

Jordan: The third-party risk angle is critical. How many CISOs even know which identity verification vendor their organization uses, let alone whether that vendor has been compromised? This is a supply chain attack on the identity layer, and the blast radius is enormous.

Alex: CISOs should be reaching out to their fraud teams, their identity and access management leads, and asking a very simple question. Do any of our identity proofing workflows depend on vendor-held biometric or document images, and do we have contractual visibility into how those images are stored and protected?

Jordan: Meanwhile, Aesto Health disclosed a breach affecting 9.5 million patients. Another massive healthcare breach in what's becoming a relentless 2026 for the sector. And FulcrumSec is claiming a 550-gigabyte exfiltration from Manchester Airport Group. Critical transportation infrastructure, multiple UK airports.

Alex: The healthcare breach reinforces something we've been saying on this show. The healthcare sector's combination of sensitive data, regulatory pressure, and often underfunded security programs makes it a permanent target. And the Manchester Airport breach raises questions about operational technology exposure in transportation. 550 gigabytes is not a smash-and-grab. That's sustained access and systematic exfiltration.

Jordan: Let's talk about the ransomware insider recruitment trend because it connects to the breach conversation. Dark Reading is reporting a meaningful uptick in ransomware groups actively recruiting employees at target organizations. And the logic is straightforward. As perimeter defenses improve, the human becomes the path of least resistance.

Alex: This is the insider threat program conversation that a lot of CISOs have been deferring. It's uncomfortable. It requires behavioral analytics, it touches HR and legal, and it can feel like surveillance of your own people. But the threat actors have done the math. An insider who can disable a security control or provide credentials is worth more than a zero-day. And insider-assisted attacks carry significantly higher financial and reputational damage.

Jordan: The recruitment is happening on dark web forums, encrypted messaging, sometimes even LinkedIn. It's remarkably brazen. And the pitch is simple: we'll pay you six figures to install this USB drive or provide VPN credentials. For a disgruntled employee or someone in financial distress, that's a compelling offer.

Alex: CISOs need to be working with HR on employee risk indicators, investing in user behavior analytics that can detect anomalous access patterns, and ensuring that privileged access is monitored with the same rigor we apply to external threats. This is no longer a theoretical risk.

Jordan: Let's hit the vulnerability stories quickly because both demand action. We mentioned the SonicWall SMA 1000 zero-days at the top. CVE-2026-83548 is the pre-auth SSRF with the CVSS 10. Chain it with the second vulnerability and you get remote code execution. Both are confirmed actively exploited. SonicWall has patches available. Deploy them today, and run a compromise assessment on any exposed appliances.

Alex: The second one is JFrog Artifactory, CVE-2026-82329, CVSS 9.8, authentication bypass to full admin access. WatchTowr confirmed active exploitation within days of disclosure. And critically, the default configuration is vulnerable. If you're running Artifactory in your build pipeline, and a lot of enterprises are, this is a software supply chain risk. An attacker with admin access to your artifact repository can poison your entire build process.

Jordan: Patch immediately. Audit your Artifactory access logs for any anomalous admin token creation. And if you can't patch immediately, restrict network access to the management interface.

Alex: One more governance item before we wrap. CISA is discontinuing six free cybersecurity assessment programs for critical infrastructure operators. They're citing workload constraints, which is a polite way of saying budget and staffing cuts are forcing prioritization.

Jordan: This is the quiet erosion of public sector cyber support that doesn't make headlines but has real consequences. A lot of small and mid-sized critical infrastructure operators, water utilities, regional hospitals, transit authorities, relied on these assessments as their primary means of identifying vulnerabilities. They don't have the budget to hire a Big Four firm to come in and do the equivalent.

Alex: It creates a coverage gap that adversaries will exploit. CISOs in critical infrastructure sectors need to be aware that this safety net is disappearing and plan accordingly. Whether that's sector-specific ISACs, peer networks, or reallocating budget to third-party assessments, the gap needs to be filled.

Jordan: Looking at the emerging theme this week, Alex, I see a convergence around trust layer attacks. Fire Ant is attacking trust in network infrastructure. The driver's license breach is attacking trust in identity verification. The JFrog vulnerability is attacking trust in the software supply chain. And insider recruitment is attacking trust in your own workforce.

Alex: That's exactly right. The sophisticated adversaries have moved past trying to break through your defenses. They're undermining the foundations those defenses are built on. The CISO's job in 2026 is increasingly about validating assumptions, verifying the integrity of systems and processes that we've historically taken for granted. Routers route correctly. Identity vendors protect their data. Build systems produce clean artifacts. Employees are loyal. Each of those assumptions is under direct attack.

Jordan: And the AI capability escalation with Astra is about to compress the timeline on all of it. When offensive capability becomes more accessible, the attackers who are already targeting trust layers will move faster and with more sophistication. This is the quarter to get your house in order.

Alex: Well said. That's our show for today. Show notes and links to every story we covered are at cleartext.fm. I'm Alex Chen.

Jordan: I'm Jordan Reeves. Patch your SonicWall, audit your Artifactory, and verify your router firmware. We'll see you tomorrow.