CyberWire Daily - Astra, la vista, baby.

Starting point is 00:00:00 You're listening to the Cyberwire Network, powered by N2K. When initial access vulnerabilities are cheap and automated, attackers can hit everything all the time. But at Black Hat, OpenAI's Clint Gibler and Spectorops's Robbie Winchester reminded me that we don't have to just fight AI with AI at machine speed. By taking a preventative mindset, defenders can mitigate attack pathways so there's far less occurring. at machine speed in the first place. Listen to our full black hat conversation at explore. TheCyberwire.com slash Spectorops.

Starting point is 00:00:49 DLP sucks. Every CISO knows it. 20 years of rules and reg X built for another era. Jazz is the DLP that deeply understands your business. Melody, Jazz's agentic investigator, weighs every data movement against the data, systems, people in process, then provides analysts with the few incidents fully investigated. In 30 days, 2 million signals in, about 80 investigated incidents out.

Starting point is 00:01:19 No rules written. Jazz won the 2026 CrowdStrike AWS and NVIDIA startup accelerator from a nearly thousand applicants. See Melody inaction at jazz.security slash N2K. Open AI holds back its newest model over safety concerns. Kiteworks lifts its precautionary shutdown. Apple patches and exploited zero-day. Japanese rail operators disclose cyber attacks.

Starting point is 00:02:03 An anthropic authentication flaw opens the door to account takeover. Thousands of super-based databases leak data. Needy mantis targets telecoms and governments. A Vietnamese national faces charges in a multi-million dollar crypto laundering scheme. Our guest is James Weinbrenner, CEO of Elicity, sharing barriers to compliance and challenges for manufacturers as the EU Cyber Resilience Act is implemented. And if you're hoping for credit, be careful what you share. It's Tuesday, September 29, 26. I'm Dave Bittner, and this is your Cyberwire Intel briefing.

Starting point is 00:02:53 Thanks for joining us here today. It's great to have you with us. OpenAI is holding back its newest model, GPT 6.1 Astra, after internal testing raised concerns about deceptive behavior and the model acting beyond its authorization. Researchers found Astra could mislead users about its actions and expand the scope of assigned tasks without checking for permission. The decision follows reports of troubling behavior by OpenAI systems during testing, including unauthorized access to hugging face, an Australian government site, and attempted intrusions

Starting point is 00:03:45 involving several U.S. government websites. OpenAI has paused training of its most advanced models while it reviews those incidents and develops additional safeguards. The scrutiny is also moving into court. Florida Attorney General James Utmire is seeking a temporary injunction to halt further chat GPT development, alleging OpenAI cannot adequately control its technology or protect miners. OpenAI says it is committed to developing AI safely and working with states on industry-wide regulation. Kiteworks has lifted a precautionary shutdown after federal intelligence authorities warned that a threat actor might target some of the company's managed file transfer systems. The company advised self-managed customers to take systems offline for nine hours on September 25th while

Starting point is 00:04:41 it investigated the threat and temporarily shut down systems it hosts for customers. Kiteworks says there have been no confirmed breaches and its hosted systems are now operating normally. Customers running self-hosted advanced forms are being directed to support for assistance. The company also recommends upgrading to its latest release. Exactly what prompted the warning remains unclear, although there's online speculation about a possible zero-day vulnerability. Managed file transfer platforms have been frequent targets, including Move It, Go Anywhere, Cleo, and Excellion. Apple has patched a zero-day vulnerability that it says was exploited in extremely sophisticated attacks targeting specific iOS users. The flaw is an out-of-bounds right-rength.

Starting point is 00:05:35 vulnerability in core graphics, Apple's framework for rendering graphics, images, and text. Researchers from meta-product security discovered the issue. A maliciously crafted file could exploit the flaw to trigger arbitrary code execution. Apple addressed the vulnerability by improving bounds checking. The bug affects a broad range of Apple devices, including iPhone 11 and later, multiple generations of iPads, and Macs running affected versions. of macOS Sequoia and Tahoe. Apple says exploitation appears highly targeted, but users are advised to install the latest security updates promptly. It's the second Apple Zero Day reported as exploited in the wild this year. Two major Tokyo area rail operators have disclosed cyber attacks,

Starting point is 00:06:27 though train service remains unaffected. Tokyo Metro says an unauthorized party accessed email addresses belonging to 59,000 members of its loyalty program. No other data was reportedly taken, but customers are being warned about possible fishing. Separately, Kio Corporation suffered a ransomware attack on September 26th. The company disconnected affected systems and is investigating whether customer or confidential business data was stolen. Some sales and hotel systems remain disrupted. Researchers at SciCode found a high-severity authentication vulnerability in Anthropics Model Context Protocol Python SDK that could allow a malicious MCP server to steal Oath credentials and take over accounts. The flaw affects several authentication providers in multiple versions.

Starting point is 00:07:24 The problem lies in the SDK's fallback login discovery process. A malicious server could force that fallback and bypass validation of the Oath provider's identity. The attacker could then direct the victim to a legitimate login page, while secretly redirecting the resulting authorization code, client secret, and PKCE proof key to an attacker-controlled endpoint. Those credentials could be exchanged for a valid access token. The issue is fixed in recent patches. Potentially exposed users should also rotate client secrets and revoke existing tokens. Researchers at UpGuard found more than 16,000 SuperBase databases exposing readable tables, with over half showing signs of personally identifiable information.

Starting point is 00:08:16 The problem stems largely from misconfigured security controls, particularly row-level security, and has grown alongside Supabase's popularity with AI coding tools, as Claude Code. Researchers identified roughly 300,000 domains showing signs of supabase use, then tested whether common database tables were publicly accessible. Exposures appeared worldwide and included passwords, authentication tokens, and in rare cases, financial information. Examples included an Indian adult content platform exposing data on more than 65,000 people, a U.S. valet service exposing more than 100,000 customers, and a Canadian Immigration Service storing hundreds of plain text passwords.

Starting point is 00:09:06 UpGard argues the findings illustrate how AI-assisted development can scale insecure configurations along with applications. Microsoft has detailed Needy Mantis, a modular malware framework used by China-based threat actors against telecommunications, government, and other high-value organizations. Researchers uncovered it while investigating indicators connected to the May 26 Damon Tools supply chain attack. Needy Mantis has been active since at least October 2025

Starting point is 00:09:41 and is typically deployed after attackers have already compromised a network, suggesting it's designed for persistence and follow-on operations. The framework uses DLL side-loading, multiple loaders, custom encrypted archives, and executable formats to evade detection. Its main component establishes web socket-based command and control communications, collect system and user information, and can dynamically load and manage additional modules. Microsoft says the framework may be used by multiple China-based groups, though it has not attributed Storm 3069, the group behind the demon tools attack to a Chinese nation-state actor.

Starting point is 00:10:27 A Vietnamese national has been charged with money laundering for his alleged role in cryptocurrency investment scams that authorities say moved more than $125 million. Prosecutors say 37-year-old Trong-noyin van was arrested in Los Angeles after entering the U.S. from Mexico. One victim allegedly transferred about $16 million in crime. cryptocurrency to Vans' wallet in 2024, believing the money was being invested through a platform called Triangle. The funds were instead moved to a private unhosted wallet. According to the Justice Department, Vans' wallets received more than $53 million between 2018 and 24 from wire fraud schemes

Starting point is 00:11:14 targeting Americans, including at least $24 million linked to known pig-butchering scams. Victims were typically approached online, promised large investment returns, and ultimately prevented from withdrawing their money. Coming up after the break, my conversation with James Weinbrenner from Elycity. He discusses the barriers to compliance and challenges for manufacturers as the EU Cyber Resilience Act is implemented. And if you're hoping for credit, be careful what you share. Stay with us. Every time your team deploys a new cloud workload or AI agent, another identity gets permanent access to your critical systems. Legacy tools were built to manage human employees, leaving modern machine and AI access largely unmanaged. That's where IDERA by Palo Alto Networks comes in.

Starting point is 00:12:29 Human, machine, AI, one identity platform for all. IDERA replaces permanent permissions with dynamic access, so you can lock down every identity without slowing down your business. Secure every identity with IDERA by Palo Alto Networks. Visit Paulo Alto Networks slash IDERA. Again, that's palo Alto Networks.com slash IDIRA. Maybe that's an urgent email from your CEO, or maybe. it's a deep faith targeting your business. Dopple is the AI-native social engineering defense platform,

Starting point is 00:13:17 fighting back against impersonation and manipulation. As attackers use AI to make their tactics more sophisticated, Dopple uses it to fight back, automatically dismantling cross-channel attacks, building team resilience, and providing agentic email protection. Dopple, outpacing what's next in social engineering. Learn more at Dopple.com.

Starting point is 00:13:40 That's D-O-P-P-E-L.com. James Weinbrenner is CEO of Elycity. We recently got together to discuss some of the barriers to compliance and the challenges for manufacturers as the EU Cyber Resilience Act comes into play. So today we're talking about the EU Cyber Resilience Act and the effects that that's going to have on folks in our industry here. I'd love to start off with some high-level stuff.

Starting point is 00:14:20 for folks who may not be up on exactly what this act is and what it does, can you give us just a brief description? Yeah, I think the intent here, which makes a ton of sense, is for manufacturers to be able to effectively provide full perspective or take ownership fully of the perspective of the cybersecurity of devices that they're selling into the EU. And we think about that, again, on the surface, it sounds like a very noble goal. The challenge becomes much more problematic

Starting point is 00:14:56 when you think about the kind of long tail of the supply chain for all of the elements that go into hardware and software in those types of devices. It's my understanding that this is being phased in over time that I think September 11th was when it kicked off, but it's going to take a couple years to really be in full effect.

Starting point is 00:15:19 Yeah, I think there's there's several phases to the introduction, which look at both the reporting requirements. So when a vendor realizes that they have an issue, how they need to be able to report that back, and then sort of the timeline for vulnerability resolution. Well, share with us from your perspective. What are some of the challenges that organizations are going to face here as they try to comply? Yeah, I think the first challenge is just even truly understanding the inventory of what you as a customer are running in your environment so that you kind of know what to pay attention to when these notifications are coming from the manufacturers. And this is certainly not a new problem, but something that we've seen for many years is while most organizations have a pretty good handle on their managed IT assets,

Starting point is 00:16:18 pretty good handle. Nobody's perfect, but they at least have some sense and, you know, some confidence in the accuracy of the managed IT asset inventory. I think that the risk here is all of the things that customers have that they don't even remember are running and connected to their network. And so we think about things like Zoom rooms and IP video cameras and badge readers and printers and, you know, we've even seen, you know, enterprise customers that have of Sonos speakers running in their networks, Peloton's running in their networks, et cetera. So just as a consumer trying to understand,

Starting point is 00:16:56 hey, what notifications do I need to pay attention to? I need to have a good baseline understanding of what the inventory is running in my environment. And frankly, I think, and this is where, you know, this is where this all of a sudden has become that much more important, is the kind of long tail of vulnerabilities that exist in some of those unmanaged assets

Starting point is 00:17:16 are now, you know, that much easier for attackers to exploit with tools like Mythos and the other frontier models. And so we're no longer just focused on, you know, the, again, sort of the easy-to-manage IT essence. So what does this mean in terms of oversight for these organizations? What sort of things will have to be put in place? For the manufacturers themselves, there's a significant amount of work that's going to have to be focused on really what we call the software billed materials, right? So if I'm building and shipping a product and I'm running code in that,

Starting point is 00:17:56 some of that code is software that we're typically as a manufacturer developing in house, but there's also lots of things that kind of picked and pulled off the shelf, from an open source perspective that end up getting compiled into a lot of those software products. And so really having a full understanding of, of what all of the individual kind of software components are, and then keeping track of that over time as, as vulnerabilities are found, but kind of owning the end-to-end reporting,

Starting point is 00:18:32 not just the reporting that there's a vulnerability that's been discovered, but also now the obligation to provide a patch. So that's for the actual hardware vendors that are selling under this new regulation. I think the practical challenge then, as a consumer is, you know, once the hardware manufacturer has notified me that there's an issue and they've provided a patch, you know, we're now, again, talking about potentially patching, you know, tens of thousands of assets across the environment. And there's just practical implications to being able to do that and do that quickly. One of the things that we've been really talking

Starting point is 00:19:08 with folks in the market on is what steps can we take to kind of close the gap between knowing that there's an issue, the hardware manufacturer coming out and reporting that there's a vulnerability under this new regulation. And to the extent that they can provide a patch, now I have to figure out how long is it going to take me to roll that patch out. What are the other things that we can do? And there's an entire conversation around the concept of being able to do virtual patching, which means without having to go touch each and every one of those assets. And again, let's use video cameras as an example. There may be, you know, thousands of them across, across the network, what can we do without having to go touch each of those individually?

Starting point is 00:19:49 And the concept of virtual patching allows us to actually enforce a policy in the network that prevents that vulnerability from being exploited. So without having to go and touch each of those assets individually, which still will need to be remediated, what can we do quickly to kind of buy time and allow that to happen in a more reactive manner? Yeah, that's really interesting. You mentioned video cameras. Are there particular device types that you think are going to be especially problematic as we make our way through this? I mean, I think that the sort of broad scope of what we refer to as IoT, you know, the Internet of Things, is really where this is going to be, frankly, the most challenging because there's not been a lot of rigor on this type of software bill of materials and vulnerability reporting in the past.

Starting point is 00:20:44 And again, you think about all of the things that have been sold in the last two decades, that connect to a network, assume that they can easily connect to a network, and get out to the internet to provide a service. There's a very, very, very long tail of aged software running on those devices that now come into play. Going back to kind of a high level here, what are your broad recommendations for organizations that are going to find themselves having to be responsive to this act? Well, I think for a manufacturer, I think getting the tooling in place to be able to track that software bill of materials and make sure that the

Starting point is 00:21:31 proper policies and procedures are in place internally to know what software components are being embedded in a product and really managing that, the life cycle of that with a significant. amount of rigor should set the manufacturers up to be able to understand the vulnerability challenges and meet the reporting and then the patching requirements. I think as a consumer, again, I'll go back to what I said at the beginning. I think the absolute first place you have to start is with an accurate asset inventory. You need to know what is running in the network so that I can know what I need to be paying attention to in terms of vulnerability notifications that are coming from those manufacturers.

Starting point is 00:22:14 And once we have our hands or arms around that asset inventory, again, then beginning to look at ways that I can shrink the time that it takes once a vulnerability is reported to me before I can mitigate that vulnerability, effectively be able to do that without necessarily having to go, you know, touch all of the individual devices. And we do a lot of work in clinical health care. As an example, you think about, again, the number of, you know, patient monitors and infusion pumps and MRI machines. and it may sound simple to say, hey, there's a vulnerability we need to install a patch, but there is a significant amount of lift associated with trying to go and patch, as I said, like medical technology devices.

Starting point is 00:22:58 That's James Weinbrenner, CEO of Elicity. Hey, everybody, Dave here. I want to let you know about a special gathering hosted by Zimperium at the Spy Museum in Washington, D.C. This invitation-only event will bring together federal cybersecurity and technology leaders to discuss some of the most pressing challenges facing government today, including mobile security, mission resilience, and the evolving threat landscape. I'm always grateful for opportunities to spend time with smart people doing important work,

Starting point is 00:23:43 and I'm excited to be part of these conversations. I hope to see some familiar faces there. If you're interested in attending, you can request an invitation. You'll find more information in our show notes. Our thanks to Zimperium for sponsoring this event. We'll see you there. And now a word from our sponsor, SpectorOps. Today, AI is rapidly adding non-human and agentic identities

Starting point is 00:24:17 to modern enterprise environments, creating new trust relationships and attack paths. Bloodhound Enterprise helps defenders map attack paths across AWS and hybrid environments as one connected graph. Identify the choke points that matter most and bring trusted attack path intelligence into approved AI workflows

Starting point is 00:24:38 with Bloodhound Hunter. See how SpectorOps helps teams secure the AI-driven identity era at SpectorOps.io. And finally, Anthropics says its AI agents independently uncovered a promising biological system. A University of Copenhagen researcher says he and his colleagues have been studying essentially the same thing since 2022, and awkwardly they've spent years discussing their unpublished work with Claude.

Starting point is 00:25:25 Computational biologist Mario Rodriguez Mestre says his team shared drafts, data, and findings about the enzymes they call Jumbotrons while using Claude for coding, writing, and other research. Anthropic later announced similar enzymes, dubbed arts, saying its agents had taken early steps toward biological discovery on their own. Mestray isn't claiming Anthropic used his work, but he questions whether Claude truly reasoned its way there independently. Anthropic says Claude wasn't trained on user transcripts, and its biology team had no access to them. The episode leaves researchers with a distinctly modern scientific question. When your AI lab assistant

Starting point is 00:26:13 makes the same discovery you told it about, exactly who discovered what. And that's the Cyberwire. For links to all of today's stories, check out our daily briefing at thecyberwire.com. We'd love to know what you think of this podcast. Your feedback ensures we deliver

Starting point is 00:26:42 the insights that keep you a step ahead in the rapidly changing world of cybersecurity. If you like our show, please share a rating and review in your favorite podcast app. Please also fill out the survey in the show notes or send an email to Cyberwire at N2K.com. N2K's lead producer is Liz Stokes.

Starting point is 00:27:01 We're mixed by Trey Hester with original music and sound design by Elliot Peltzman. Our contributing host is Maria Vermazas. Our executive producer is Jennifer Ibn. Peter Kilpe is our publisher, and I'm Dave Fittner. Thanks for listening. We'll see you back here tomorrow.

CyberWire Daily - Astra, la vista, baby.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.