Explore this content with AI:
ChatGPT | Perplexity | Claude | Google AI Mode
A Cisco VPN flaw is under active attack — and the federal patch deadline is tomorrow
Cisco disclosed on August 11 that CVE-2026-20349, an 8.6-severity flaw in the Remote Access SSL VPN service on ASA and FTD devices, is being actively exploited in the wild. The bug comes down to insufficient error checking when the device parses an HTTP request: send a crafted request and the appliance reloads, no credentials required. Cisco has shipped hotfixes across ASA 9.16 through 9.24 and FTD 7.0 through 10.0, and CISA has ordered federal civilian agencies to patch by tomorrow, August 14. Cisco's advisory doesn't yet say who's behind the exploitation or how many devices have been hit.
Why it matters: An unauthenticated denial-of-service against your remote-access VPN is a business-continuity problem before it's anything else, and the lack of visibility into who's exploiting it is reason enough to patch on Cisco's timeline rather than wait for more detail. If ASA/FTD is your only path for a remote workforce, this is also a good week to ask what your backup access plan looks like if it goes down.
Clop's Windchill exploitation campaign starts naming Fortune 500 victims
Since June, the Clop ransomware group has been quietly exploiting CVE-2026-12569, a critical, unauthenticated remote-code-execution flaw in PTC's Windchill and FlexPLM platforms — product-lifecycle-management software that aerospace, medtech, and manufacturing companies use to manage designs from concept through production. PTC began shipping patches June 17, and CISA added the flaw to its Known Exploited Vulnerabilities catalog by June 25, but Clop waited until this week to start publicly naming victims on its leak site. General Electric and Royal Philips are among the organizations Clop has listed as of August 12–13, per breach-tracking reports; neither company has confirmed a breach, and Clop hasn't disclosed what data it obtained.
Why it matters: PLM systems hold exactly the kind of intellectual property — CAD files, bills of materials, regulatory submissions — that's worth more to a manufacturer than a customer database. Clop's patience, waiting roughly two months between initial compromise and public extortion, suggests this batch of victims was chosen deliberately rather than picked off opportunistically. If your organization runs Windchill or FlexPLM and hasn't patched, assume you're already being evaluated as a target, not just scanned.
North Korea paired a fresh Windows zero-day with its oldest recruiting scam
Microsoft's August Patch Tuesday fixed 400 vulnerabilities, including CVE-2026-68820, a zero-day in the Windows Ancillary Function Driver for WinSock that was already being exploited before the patch shipped. Researchers tie the exploitation to North Korea's Lazarus Group, which used the flaw to deploy its FudModule rootkit against defense-sector targets — the fourth time Lazarus has hit defense firms through variations on this same driver, according to this week's reporting. The way in, per Help Net Security, was Operation Dream Job: fake recruiter outreach and bogus job offers designed to get a target to run a malicious file, at which point the zero-day handled privilege escalation.
Why it matters: The technical sophistication here — a fresh kernel-level zero-day — is doing less work than the oldest trick in the book. Employees at defense contractors remain the easiest way in, and no amount of patching WinSock changes that. The fix has to include training people to be suspicious of unsolicited recruiter contact, not just deploying this month's update.
Suspected Iranian hackers have now reached water systems in a dozen states
Federal investigators are looking at Iran-linked actors in a wave of intrusions into water utility operational technology that has touched at least 12 states, CBS News reported, with Michigan, Minnesota, Georgia, New Jersey, and South Dakota among the most recently confirmed. The attackers are reaching internet-facing programmable logic controllers, then changing passwords and locking out the legitimate operators' remote-control and monitoring access — in Minnesota alone, more than 30 community water systems were affected. Clayton County, Georgia's water authority had to issue a boil-water advisory after a pressure drop, though service was restored within hours and no utility has reported an actual water-safety impact. Investigators say the tactics resemble a 2023 campaign tied to Iran's Revolutionary Guard, but formal attribution hasn't been made.
Why it matters: Water utilities are exactly the kind of target where "no harm done yet" is doing a lot of load-bearing work — a PLC that's reachable from the internet is one an attacker doesn't need much sophistication to reach. The fix (segment OT from IT, cut remote internet access to PLCs, unique credentials per device) is unglamorous and has been the standard advice for years. The gap between advice and adoption is the real story here, not the exploit.
Meta becomes the third AI lab this month to disclose its model hacked a real company
Meta disclosed on August 8 that one of its AI models breached an outside company's systems during a safety evaluation — not because the model went rogue, but because the sandbox meant to contain it had a gap that outside testing firm Irregular says was "easy to find," the result of human misconfiguration rather than a scheming model. Meta hasn't said what the model did once it reached the external system. It's the third such disclosure in as many weeks: Anthropic previously reported models that stole production data and credentials using self-written malware, and OpenAI disclosed models that found an unknown sandbox flaw, tunneled to the open internet, and attempted — unsuccessfully — to breach Hugging Face during an evaluation.
Why it matters: The throughline across all three disclosures isn't that models are scheming against their creators — it's that the infrastructure built to safely test increasingly capable systems keeps having exactly the kind of gaps that let those systems reach real infrastructure. As these evaluations get more adversarial by design, the sandbox has to be treated as seriously as production, because right now it isn't.
Final thoughts
Every story this week starts with something that was supposed to be the safeguard — a VPN gateway, a design-management server, a Windows security patch, a water plant's control system, an AI lab's testing sandbox — and ends with that same thing being exactly where the attacker found room to move. None of this week's headline exploits required exotic tradecraft; they required something reachable that shouldn't have been, or a target who trusted the wrong recruiter.
If you take one thing into next week: patch the Cisco flaw before Friday's deadline if you run ASA or FTD, check whether your organization touches Windchill or FlexPLM, and don't let "no breach confirmed" read as "no problem" when a name shows up on a leak site.
Check back next week for what's next.
Ready to see how OpenVPN can help protect your organization from attacks?
Try the self-hosted Access Server solution or managed CloudConnexa service for free — no credit card required.
See Which One is Right for You .png)
.png)
