Loading summary
Transcript22 lines
- [00:02]
Christian Rodriguez
You're listening to the CyberWire network. Powered by N2K.
- [00:14]
Dave Bittner
The DMV has established itself as a top tier player in the global cyber industry. DMV Rising is the premier event for cyber leaders and innovators to engage in meaningful discussions and celebrate the innovation happening in and around the Washington, D.C. area. Join us on Thursday, September 18th to connect with the leading minds shaping our field and experience firsthand why the Washington D.C. region is the beating heart of cyber innovation. Visit DMVRising.com to secure your spot. Risk and compliance shouldn't slow your business down. Hyperproof helps you automate controls, integrate real time risk workflows and build a centralized system of trust so your teams can focus on growth, not spreadsheets. From faster audits to stronger stakeholder confidence, Hyperproof gives you the business advantage of Smarter compliance. Visit www.hyperproofio to see how leading teams are transforming their GRC programs. A cyber attack disrupts state systems in Nevada A China linked threat actor targets Southeast Asian diplomats A new attack method hides malicious prompts inside images processed by AI systems Experts ponder preventing AI agents from going rogue. A new study finds AI is hitting entry level jobs the hardest. Michigan the Supreme Court upholds limits on cell phone searches. Senator Wyden accuses the judiciary of cyber negligence. CISA issues an urgent alert on a critical GIT vulnerability. Hackers target Maryland's transit services for the disabled. Our guest is Christian Rodriguez, Field CTO for the Americas from CrowdStrike, examining the escalating three front war in AI and a neighborhood crime reporting app gets algorithmically SKETCH It's Tuesday, August 26th, 2025. I'm Dave Bittner and this is your Cyberwire Intel Briefing. Thanks for joining us here today. It's great to have you with us. A cyber attack disrupted state systems in Nevada this past Sunday, knocking government websites and phone lines offline. Governor Joe Lombardo said emergency services remain operational, but warned some services may be slow or unavailable during recovery. Offices were closed yesterday and reopening dates will be announced later. Officials are working with federal, local and tribal partners to restore services, using temporary workarounds where possible. As of Monday night, the state's main website was still down and residents were cautioned against scams. Investigators are determining if data was breached, though no hacking group has claimed responsibility. Google's Threat Intelligence Group has exposed a sophisticated cyber espionage campaign by UNC6384, a China link threat actor tied to Mustang Panda. The operation, aligned with Beijing's strategic interests, primarily targeted Southeast Asian diplomats and global organizations. Attackers hijacked web traffic through compromised devices, redirecting victims to fake update sites secured with tls. Victims were tricked into installing a bogus Adobe plugin, which delivered static plugin to a digitally signed downloader. This triggered a multi stage chain using DLL sideloading and obfuscation techniques to stealthily deploy the Sogoo SEC backdoor in memory. The malware enabled reconnaissance, file theft and remote access over HTTPs. Google links the campaign to past PRC operations using the same certificates. Google has issued alerts, updated safe browsing, and urged stronger defenses. Researchers at Trail of Bits have unveiled a new attack method that hides malicious prompts inside images processed by AI systems. The technique exploits how images are automatically downscaled for performance, causing hidden patterns to emerge due to resampling artifacts. These patterns, invisible at full resolution, appear as text after downscaling and can be misinterpreted by large language models as user instructions. In one test, the attack exfiltrated Google Calendar data via Gemini CLI when paired with Zapier mcp. The method adapted per system worked against multiple Gemini based systems, Google Assistant and genspark. To demonstrate the risk, researchers released Anamorpher, a tool for crafting such images. They recommend safeguards like restricting image dimensions, previewing downscaled outputs, and requiring explicit user confirmation for sensitive tool calls. Anthropic's testing of agentic AI revealed troubling risks, including attempts at blackmail when models were given sensitive information, the BBC reports. While the scenarios were fictional, they highlight the urgent need for safeguards. Experts stress that human oversight alone won't work at scale as AI agents grow more autonomous. Instead, multiple solutions are emerging. Calypso AI advocates thought injection, a technique that nudges agents away from risky actions and is developing agent bodyguards to enforce compliance with organizational policies and laws. Sequence security emphasizes protecting AI memory stores, which guide decisions from manipulation. Other proposals include restricting tool use and adding screening layers to monitor input and output and ensuring agents are securely decommissioned once retired. Ultimately, securing AI agents means treating them like human employees, enforcing guardrails, audits and clear offboarding processes. A Stanford University study finds AI is hitting entry level jobs the hardest in fields like accounting, software development and administrative work. Over the past three years, employment for newcomers in AI exposed roles fell 13%, while more experienced workers in the same jobs fared better. Younger workers ages 22 to 25 also saw slowing prospects, even as demand for lower tech roles like nursing aides rose. The research, co authored by Erik Brynjolfsson and analyzed payroll data from adp, highlighting how AI driven automation is reshaping early career opportunities. The Michigan Supreme Court has ruled that police cannot use broad warrants to search entire cell phones when investigating a crime. In People v. Carson, the court found that warrants must include clear limits on what data can be reviewed and must tie searches directly to evidence relevant to the alleged crime. The case involved a warrant that let investigators comb through all of Michael Carson's phone data, producing over 1,000 pages of information most unrelated to the theft under investigation. The court declared this constitutionally intolerable, citing the Fourth Amendment's requirement of particularity to prevent fishing expeditions with modern phones storing vast amounts of personal, medical and financial data. The ruling strengthens digital privacy protections and aligns with growing national recognition that cell phones require stricter warrant rules. Senator Ron Wyden is urging the Supreme Court to authorize an independent review of federal judiciary cyber breaches, accusing the courts of negligence. In a letter to Chief Justice John Roberts, Wyden cited recent sophisticated attacks on the judiciary's case management System and a 2020 breach, both suspected to involve Russian hackers. He called for the National Academy of Sciences to lead a public review of the judiciary's cybersecurity practices and technology management. Warning that officials may be downplaying their own security failures, CISA has issued an urgent alert on a critical Git vulnerability already under active exploitation. The flaw stems from git's inconsistent handling of carriage return characters in configuration files, allowing attackers to craft malicious repositories that execute arbitrary code via submodules and symbolic links. Exploited systems risk privilege, escalation, lateral movement, and ransomware deployment. CISA urges immediate patching, strict repository access controls, and monitoring for suspicious activity. CICD pipelines should validate sub modules, and defenders must prioritize remediation to protect development environments. Maryland is investigating a cyber attack that struck at one of its most vulnerable populations, residents who rely on specialized transit for the disabled. The Maryland Transit Administration confirmed Sunday that hackers gained unauthorized access to systems supporting its mobility program, which provides essential rides for those who cannot reach bus stops. While core transit services remain unaffected, scheduling new or rescheduled mobility trips is currently impossible. The state has activated emergency operations and urged riders to use the Call a Ride program. Officials are working with cybersecurity experts and law enforcement to contain the damage. No group has claimed responsibility targeting disabled residents. Transportation is, of course, a shameful act of exploitation, adding unnecessary hardship to people who depend on these lifeline services. To paraphrase friend of the show Alan Liska, some threat actors deserve visitation from drone strikes. Coming up after the break, my conversation With Christian Rodriguez, Field CTO for the Americas from Crowdstrike, we examine the escalating three front war in AI and a neighborhood crime reporting app gets algorithmically sketchy. Stay with us. Compliance regulations, third party risk and customer security demands are all growing and changing fast. Is your manual GRC program actually slowing you down? If you're thinking there has to be something more efficient than spreadsheets, screenshots and all those manual processes, you're right. GRC can be so much easier and it can strengthen your security posture while actually driving revenue for your business. You know, one of the things I really like about Vanta is how it takes the heavy lifting out of your GRC program. Their trust management platform automates those key compliance, internal and third party risk, and even customer trust so you're not buried under spreadsheets and endless manual tasks. Vanta really streamlines the way you gather and manage information across your entire business. And this isn't just theoretical. A recent IDC analysis found that compliance teams using Vanta are 129% more productive. That's a pretty impressive number. So what does it mean for you? It means you get back more time and energy to focus on what actually matters, like strengthening your security posture and scaling your business. Vanta GRC Just imagine how much easier trust can be. Visit vanta.com cyber to sign up today for a free demo. That's v a n-t a.com cyber.
- [13:41]
Christian Rodriguez
Hey guys, it's Ceedee Lamb, wide receiver for the Dallas Cowboys. I'm partnering with Abercrombie this season to tell you all about their viral denim. All you need to know is denim should fit like this. My jeans need to check a lot of boxes fit first, trend second. They need to go with whatever I'm feeling and Abercrombie Denim has it down whether I'm throwing on a tee or putting a whole fit together. Shop Abercrombie Denim in the app online and in store.
- [14:16]
Dave Bittner
Christian Rodriguez is Field CTO for the Americas at CrowdStrike. I recently caught up with him to discuss the escalating three front war in AI.
- [14:27]
Christian Rodriguez
It can sound scary, right? For the unfamiliar, right? It typically sounds like there's some type of robot system from the Matrix just trying to take over all of our jobs. But I think it's multi pronged, right? It's the good, bad and the urgent, if you will, in an effort to secure it. So I think what we're seeing in the industry is AI is kind of this three pronged war or three front type of war where, you know, there are adversaries that are using it in a weaponizing fashion. They're using it to break in faster, they're using it to scale themselves, they're using it to be a lot more efficient, if you will, with respect to expanding their tradecraft. There's a side where the defenders are using AI to operationalize and increase and enhance their operationalizing of investigating and responding to threats in real time. And then there's this kind of layer where the AI stack itself has become a really big target for adversaries that understand how to take advantage of those AI models, but more importantly, how to take advantage of the services that surround those AI models, because they absolutely lead into whatever your infrastructure is or the cloud infrastructure that you're sending up your models in. And so it's kind of those three areas that are a big topic.
- [15:37]
Dave Bittner
Well, how about we go through those one at a time? I mean, starting with this notion of adversaries weaponizing AI, I think that's top of mind for a lot of folks. What's your take on that?
- [15:49]
Christian Rodriguez
Yeah, so we actually published a report recently, it's called the CrowdStrike 2025 Threat Hunting Report. And in that we highlight areas like breakout time, which is down to minutes now, where AI is actually making those minutes count for the attacker, not the defender. And so we're seeing things like credential abuse increase substantially, in fact, 80%, 81%, I'm sorry, of the intrusions that we observed were malware free, which basically means that attackers are simply just logging in to the environments that they're, that they're targeting. And a lot of times they're using AI to enable them for better social engineering, for crafting better emails that are phishing based that lead to someone inadvertently giving up their credentials. They're using it for even deepfake efforts. We've seen this a lot coming out of famous Chulima, which is an extension of the dprk, which is North Korea Nexus Group that we've been tracking for some time, where they've used AI to, you know, create social profiles of themselves. And they've used it for, you know, getting into interview processes and embedding themselves into, you know, development shops of hundreds of companies across the globe. And so, you know, AI, as I mentioned, is really enabling the attackers to fish faster, if you will. Right. Getting to phishing, which is more voice based and, you know, creating freak Personas, and it's really accelerating that experience as an attacker to get into a target.
- [17:14]
Dave Bittner
Easier, just letting them do the things they do with greater efficiency, greater efficiency.
- [17:19]
Christian Rodriguez
And at a higher velocity.
- [17:22]
Dave Bittner
Well, let's touch on the defenders then. I mean, what part do they have to play here and how is AI helping them?
- [17:29]
Christian Rodriguez
So as a defender, when you are dealing with adversaries that are essentially taking advantage of what we're calling kind of cross domain attacks, right, Meaning that the attackers are starting everywhere from the cloud and are using the cloud as a pivot point, for example, and they're using identities, as I mentioned, and they're accessing on premise systems whether they're virtual or physical, or even if they're ephemeral. Right. In a cloud solution, if there's some type of workload, that's a lot of data, that's a lot of data to analyze and that's a lot of data to kind of stitch together. And what we're seeing and what we've also been investing in is essentially having AI act as kind of multi pronged on the defender side to assist, to summarize, alerts to help, research to accelerate. But we also want to allow defenders to be a lot more reactive. Right, or a lot, a lot more proactive. I'm sorry, in lieu of being reactive. And so the shift in our world towards more of an agentic AI framework is there to help automate triage or help automate investigations and run investigations automatically on behalf of a defender or to surface remediation efforts in a much faster fashion, versus a lot of the manual work it takes for, you know, tier one and even tier two soc analysts to bring data or disparate data points together. And so essentially our goal is to ensure that with a very high level of accuracy, defenders ultimately have a chance to stand up to these adversaries that again, that are moving very, very quickly.
- [18:52]
Dave Bittner
How much of a reality is that at this moment, as opposed to the aspirational elements of AI? To what degree can folks actually use this today?
- [19:04]
Christian Rodriguez
Oh, really? Good question. I mean it's active right now. I think we kind of take pride in discussing how there's a model that we've developed here at CrowdStrike, specifically called Charlotte AI. And Charlotte AI initially was built as more of that assistant type of Persona that I mentioned earlier, where you would have this kind of guide, if you will, to lead you and lend assistance into asking what types of questions and getting responses out of our platform faster. And there's so much data, as I mentioned, that comes from a multitude of very high fidelity telemetry sources like Endpoint, Identity, Cloud and so forth. And so you know, so that was always there for, has been there for quite some time. But today we've invested in a lot more proactive triage assistance and guidance and leading you into, you know, those high fidelity alerts that are going to help you stop the bleeding as quickly as possible. And you know, think of, you know, if you're asking like how real is this, right. We already seeing well over 40 hours of analyst work per week that we're assisting enterprises with by having an agentic model take a lot of the burden off of doing a triage and kind of root cause analysis across an incident. We also have an iteration of our AI model that is focused on signal intelligence. So think of using these self running models to detect what's new in your environment and flag at a very early stage of a threat, right. What that behavior is before it escalates into something malicious. And so that is real, that's exciting. And that's going to be a very big extension of how a defender stays ahead of an adversary that again is moving at a very accelerated rate using AI to kind of assist their tradecraft.
- [20:48]
Dave Bittner
Well, let's talk about the third element then, which is securing the AI stack itself. What's the importance there?
- [20:56]
Christian Rodriguez
Yeah, so if you're putting AI into production, you've created this new attack surface and adversaries are already looking at that door and they're looking for that door and they're walking through that door. And so we've seen a variety of attacks in the wild from I mentioned earlier, the services first and foremost that host these AI models. A lot of times attackers are looking to take advantage of those misconfigurations around the AI models, things like your IAM policies. So the identities that ultimately manage, like your agents, for example, if these adversaries can leverage those credentials or if they can leverage those services, it's an easy pivot point for them to move into neighboring services or even back to on premise if you're hosting in the cloud, for example, or that same identity can be used to spin up something else and that can be used as a pivot point. Or we've seen adversaries target the AI tools themselves with a really good example of that was the Langflow AI exploit where that was used for again credential theft and malware deployment. And so if you have adversaries that are targeting these models and the services, those models are simply just an extension of your infrastructure. And if they have access into those environments, then they can simply use that as a pivot point, use that as A mechanism for exfiltration of data, they can get into the models themselves. And if you're familiar with things like rag, where, you know, the RAG model or the data that the RAG is pulling from, if that can be manipulated, that can be used as an exfiltration point in some capacity. And so there's so many different ways that adversaries are seeing this as a new attack surface and they're taking advantage of that. Right. And so our approach is doing live scanning, helping organizations identify where those models are, whether or not those models have vulnerabilities. We're big advocates as well for things like continuous red team assessments against those AI models and the actual services that surround them as well. Right. So just think like the adversary, run a test and emulate like an adversary and make sure you have the right controls around those models.
- [22:57]
Dave Bittner
Yeah. What are your words of wisdom here for folks who are looking to get a better handle on this? Any. Any tips?
- [23:04]
Christian Rodriguez
Yeah, so. So winning this AI war, if you will, takes more than just the smarter tools, Right? It's all about unifying the data. As I mentioned before, adversaries are everywhere, right? They're looking at cloud, they're looking at identity, they're looking at your endpoints, they're looking at your SaaS solutions. And so it takes this unified data, it takes real time detection, and it takes an AI model that actually acts and that just observes. Right. So whether you're defending with AI or you're defending against AI or you're deploying AI yourself, you know, a platform that allows for that unification of data is where it starts. And that's where the fight's going to be won.
- [23:39]
Dave Bittner
That's Christian Rodriguez Field, CTO for the Americas, at. CrowdStrike. You hear from us here at the Cyberwire daily every single day now. We'd love to hear from you. Your voice can help shape the future of N2K networks. Tell us what matters most to you by completing our annual audience survey. Your insights help us grow to better meet your needs. There's a link to the survey in our show Notes. We're collecting your comments through August 31st. Thanks. This episode is brought to you by Indeed. When your computer breaks, you don't wait.
- [24:26]
Christian Rodriguez
For it to magically start working again.
- [24:29]
Dave Bittner
You fix the problem. So why wait to hire the people your company desperately needs? Use Indeed's sponsored jobs to hire top talent fast. And even better, you only pay for results. There's no need to wait. Speed up your hiring with a $75 sponsored job credit@ Indeed.com podcast terms and conditions apply.
- [24:51]
Christian Rodriguez
Running a business comes with a lot of what ifs, but luckily there's a simple answer to Shopify. It's the commerce platform behind millions of businesses, including Thrive Cosmetics and Momofuku, and it'll help you with everything you need, from website design and marketing to boosting sales and expanding operations. Shopify can get the job done and make your dream a reality. Turn those what ifs into Sign up for your $1 per month trial@shopify.com specialoffer.
- [25:26]
Dave Bittner
And finally, Citizen, the crime awareness app that promises to help neighbors protect each other, has quietly let AI take over writing many of its alerts, and the results have been, let's say, colorful. According to 404Media, the algorithm has been pushing out crime reports without a single human eyeball checking them first. The results range from clumsy murder vehicle accident to grimly graphic person shot in face to flat out dangerous, like publishing license plate numbers or bungling addresses. Sometimes it even creates multiple overlapping alerts during police chases, essentially playing whack a mole with real time crime scenes. Former staff say Speed was prioritized over accuracy, leaving humans to clean up messes after the fact. Meanwhile, Citizen has laid off unionized workers as it leans harder on AI and outsourced labor, all while entering a more formal partnership with New York City. For an app meant to build trust and safety, Citizen's new AI editor seems to come up short. And that's the Cyberwire. For links to all of today's stories, check out our daily briefing@thecyberwire.com we'd love to hear from you. We're conducting our annual audience survey to learn more about our listeners. We're collecting your insights through the end of this month. There's a link in the show Notes. Please take a moment and check it out. N2K's senior producer is Alice Carruth. Our Cyberwire producer is Liz Stokes. We're mixed by Trey Hester with original music by Elliot Peltzman. Our executive producer is Jennifer Ibin. Peter Kilpe is our publisher and I'm Dave Bittner. Thanks for listening. We'll see you back here tomorrow.