CyberWire Daily - Meta gets a Meta-sized bill.

Episode Date: August 27, 2026

Meta settles. Australian police arrest two alleged TeamPCP members. The White House moves to shore up water utility cybersecurity. ATF reports a major cyber incident. The Navy tells sailors to lock down social media. The FBI warns of a prolific Chinese hacking operation. Bill Gates sounds the alarm on AI. A purported think tank tries to influence chatbot answers. And attackers focus less on individual vulnerabilities and more on the vendors behind them. Our guest is Tim Springston, Principal Product Manager at Semperis, on achieving hybrid identity resilience in the age of agentic AI. Meta pumps the brakes on going AI native.  Remember to leave us a 5-star rating and review in your favorite podcast app. Miss an episode? Sign-up for our daily intelligence roundup, Daily Briefing, and you’ll never miss a beat. And be sure to follow CyberWire Daily on LinkedIn. CyberWire Guest On today’s industry voices segment, we are joined by Tim Springston, Principal Product Manager at Semperis, discussing how to achieve hybrid identity resilience in the age of agentic AI. If you enjoyed this conversation, check out the full interview here. Selected Reading Meta agrees to pay $18 billion to settle US lawsuits over children's social media addiction (Reuters) Two Alleged ‘TeamPCP’ Hackers Arrested in Australia (Krebs on Security) White House to unveil program to protect water systems against hackers (POLITICO) DOJ firearms agency says hackers breached system containing investigation targets (The Record) US Navy tells sailors and their families: scrub your social media, enemies are watching (Bitdefender) Chinese Hacker Group QTFY Uses Custom-Built Platforms to Target US Infrastructure, FBI Warns (Infosecurity Magazine) Bill Gates diagnoses problems with AI, but an expert questions his prescription (ABC News) Fake US thinktank set up and funded by Israel sought to game AI for propaganda (The Guardian) SentinelOne and Tenable Find Cyber Attackers Routinely Target Edge-Device Vendor Ecosystems Rather Than Individual Vulnerabilities (SentinelOne) AI agents meant to replace Meta workers made “large-scale, disruptive actions” (Ars Technica) Share your feedback. What do you think about CyberWire Daily? Please take a few minutes to share your thoughts with us by completing our brief listener survey. Thank you for helping us continue to improve our show. Want to hear your company in the show? N2K CyberWire helps you reach the industry’s most influential leaders and operators, while building visibility, authority, and connectivity across the cybersecurity community. Learn more at sponsor.thecyberwire.com. The CyberWire is a production of N2K Networks, your source for strategic workforce intelligence. © N2K Networks, Inc.

You're listening to the Cyberwire Network, powered by N2K. What's the one thing in business that's spreading as fast as AI? AI risk. Every new tool your team signs up for, every vendor that turns on AI features, every new integration, each one is an opportunity for something to go wrong. And most security programs weren't built for AI's pace of growth. Enter Vanta. Vanta is the number one agentic trust platform.

form used by over 16,000 fast-moving companies like Ramp, Hercer, and Harvey to ensure they're always audit-ready. And now Vanta is helping companies like yours watch for the risks that show up between audits across your vendors, your AI tools, and your whole environment. The Vanta agent works like a 24-7 GRC engineer in the background, finding issues, drafting fixes for you, and cutting vendor assessment time by up to 50%. Whether you're a fast-growing startup or a global enterprise, Vanta is here to help you automate your security and compliance and earn and prove trust. Get started today at vanta.com slash cyber.

That's V-A-N-T-A dot com slash cyber. Meta settles. Australian police arrest two alleged teen PCP members. The White House moves to shore up water utility. cybersecurity. ATF reports a major cyber incident. The Navy tells sailors to lock down social media. The FBI warns of a prolific Chinese hacking operation. Bill Gates sounds the alarm on AI. A purported think tank tries to influence chatbots. Attackers focus less on individual vulnerabilities and more on the vendors behind them. Our guest is Tim Springston, principal product manager at Sempris,

on achieving hybrid identity resilience in the age of a agentic AI. And Meta pumps the brakes on going AI native. It's Thursday, August 27th, 2026. I'm Dave Bittner, and this is your Cyberwire Intel briefing. Thanks for joining us here today. It's great as always to have you with us. Meta has agreed to pay up to $18 billion over the next decade and impose new restrictions on teenage users of Facebook and Instagram. settling claims from nearly all U.S. states that its platforms were designed to addict children and harmed their mental health. The agreement limits teens to two hours of daily use and blocks access from midnight to 6 a.m. without parental consent. Meta will also restrict

most push notifications during school hours and strengthen protections against age-restricted content. The settlement includes about $12.7 billion, and guaranteed payments, with another $5 billion contingent on Snapchat, TikTok, and YouTube adopting similar safeguards. Meta also agreed to pay $459 million to resolve privacy claims stemming from the Cambridge Analytica scandal. Meta denied wrongdoing. The settlement ends of federal trial but leaves thousands of other lawsuits pending, including claims from individuals, schools, and governments, alleging social media platforms contributed to a youth mental health crisis. Australian federal police have arrested two Western Australian men, age 21 and 23, accused of involvement

with Team PCP, the Cybercrime Group linked to an extensive campaign of software supply chain attacks and data extortion. Team PCP gained prominence in late 2025 by compromising using open-source software packages and using the self-propagated shy halude worm to steal developer credentials and spread malicious code into additional projects. Its campaigns included a compromise of the Light-LLM AI Gateway that researchers say exposed secrets for more than 2,500 organizations, as well as thousands of GitHub repositories. Krebbs-on security identifies one arrested suspect as TeamPC spokesperson Ruben Thompson, based on extensive open-source research linking his online aliases to his real identity.

Team PCP appears to operate less like a traditional criminal organization and more like a loose community of collaborating hackers. The two defendants face a combined 14 cybercrime offenses. Researchers say Team PCP's attacks also help push GitHub and other software ecosystems to strengthen supply chain security. Politico reports the Trump administration is preparing a program offering free cybersecurity services to vulnerable U.S. water utilities, following a surge of attacks that officials suspect may be linked to Iran. Led by the Office of the National Cyber Director, the initiative is expected to begin as a proof of

concept in Texas before potentially expanding to other states. Utilities will receive services including network vulnerability scanning, with private cybersecurity companies reportedly helping provide the assistance. The effort comes as water systems in at least a dozen states have faced attacks affecting IT and operational technology, sometimes forcing operators to use manual controls. Federal agencies recently warned of a significant increase in attacks targeting programmable logic controllers at water and wastewater facilities. The administration has not formally attributed the campaign to Iran. The Bureau of Alcohol, Tobacco, Firearms, and Explosives says hackers breached a standalone computer system containing information about targets of ATF investigations. The agency has designated the incident, a major incident under federal guidelines. The Keelan Ransomware Group added ATF to its

leak site Wednesday, but did not publish samples of allegedly stolen data. ATF says the compromised system was isolated from its case management, laboratory, and E-Forms systems, and was quickly shut down after the breach was discovered. Officials say the attack did not affect the agency's operations. The Justice Department is investigating, while ATF has begun forensic and incident response work, Keelan remains a prolific ransomware operation, with researchers ranking it as the second most active ransomware group in July with 127 reported attacks. The U.S. Navy is warning its entire workforce, roughly 608,000 active duty personnel, reservists, and civilian employees,

to tighten their social media security amid what it describes as a coordinated intelligence gathering campaign by adversaries. In a new bulletin, the Navy advises personnel and their families to enable privacy settings and remove information linking them to the service or revealing locations and routines that could be exploited. Personnel are also being told to report suspicious activity, including surveillance of installations, drones near bases, questions about ship movements, being followed off base, and fake social media accounts impersonating systems. sailors. The warning comes amid the U.S.-Israeli military campaign against Iran. While some commenters have questioned its timing, the Navy says adversaries are attempting to gather intelligence,

test defenses, disrupt operations, and intimidate personnel. The FBI is warning that a sophisticated Chinese hacking group known as QTFI has spent years targeting U.S. government agencies and critical infrastructure, including defense contractors, communications providers, financial institutions, and universities. The group uses a custom ecosystem of tools, including QScan, a platform for identifying and exploiting vulnerable systems, and QT router, which routes malicious traffic through compromised IOT devices to obscure its origins. In 2024, QTFI reportedly exfiltrated data for more than 300 organizations,

after exploiting a checkpoint quantum gateway vulnerability. The FBI attributes QTFI to a Chinese company linked to PRC cyber operations. U.S. authorities announced Wednesday that they had disrupted Q-Scan and QT router. Agencies are urging organizations to patch systems, isolate critical infrastructure from edge devices, hunt for compromise, and audit applications for exposed credentials. Microsoft co-founder Bill Gates is warning that artificial intelligence could usher in one of the most turbulent periods in human history, unless governments act quickly to manage its risks. In a new essay, Gates argues that rapidly improving AI could replace humans across many tasks,

causing widespread job losses, economic disruption, and greater inequality. Gates says global efforts to prepare for the transition are inadequate, and suggests policies including taxes on AI usage and robots that replace workers. Revenue could support retraining programs and stronger safety nets. He also proposes reserving some roles, including caregiving and certain medical work, for humans. University of Washington computer science professor Oran Etzioni agreed with Gates' concerns, but questioned some proposed solutions, particularly taxing AI tokens. He warned that slowing U.S. development could also put American companies at a disadvantage against foreign competitors.

A website presenting itself as the Hanover Institute for Public Policy has published more than half a million words of pro-Israel content designed in part to influence how AI chatbots answer questions about Israel and Gaza, according to the Guardian. Their purported think tank has no apparent legal. entity, physical address, named staff, or report bylines. Its articles resemble neutral academic research, but consistently frame disputed issues in ways favorable to Israel. The site was built using technology marketed to help content get cited by AI systems. Piro Incorporated, the U.S. company behind the site, registered the material under the Foreign Agents Registration Act as work for the Israeli government.

The effort is part of a broader multi-million dollar public diplomacy campaign involving several contractors. Researchers warned such techniques could influence not only chat-bought search results, but eventually AI training data, potentially shaping responses without users knowing the original source.

Joint research from Sentinel 1 and Tenable suggests attackers are concentrating on particular technology vendors and product lines rather than individual vulnerabilities. Comparing exposure and runtime detection data, the companies found a 79% overlap in targeted edge device vendor surfaces, but only 21% overlap among specific vulnerabilities. The researchers argue that defenders should prioritize what Tenable calls persistently targeted vendors, product lines that repeatedly attract attackers, even as individual CVEs change. Twelve vulnerabilities were exploited independently by both state-sponsored and ransomware actors,

while more than half of organizations using F5 products had at least one exposed, actively exploited vulnerability. Citrix customers had the slowest medium remediation time at 461 days. The findings suggest organizations should combine. vulnerability remediation with a tax surface reduction and behavioral detection, particularly as AI accelerates vulnerability discovery and exploit development. Coming up after the break, my conversation with Tim Springston, principal product manager at Sempris, on achieving hybrid identity resilience in the age of agentic AI. And meta pumps the brakes on going AI native.

Stay with us. AI is transforming the way organizations work, but what happens when we rely on it so much that we begin losing the human judgment and context that make good decisions possible? I recently sat down with Johnny Hand from Trend AI, and he made an important point about what we risk when we offload too much AI.

We risk our most valuable resource, which is our human context, our creativity, our ability to understand contextually, like in the environment, those things. Those are really hard challenges for AI to tackle. If you're trying to separate AI hype from operational reality, I think you'll really enjoy this conversation. Listen now at explore.thecyberwire.com slash trend AI.

Tim Springston is principal product manager at Sempris, and in today's sponsored industry voices conversation, we discuss achieving hybrid identity resilience in the age of agentic AI. Hybrid identity is, it's not a new concept. It's been around as long as on-premise identity has been. Active Directory in 2000 was the start of it. It became hybrid from Active Directory, typically, when Microsoft and others started to have this concept of cloud.

And in order to gain access to cloud resources via Office, 365 and other services, you had to be able to sign in and you wanted to sign in with just a single identity. Well, you already had the one on premise and active directory. Why don't we have it be hybrid? And so you ended up in this hybrid scenario where you have a copy of the same identity and the same user sign in and the same access controls across two different identity stores. That's what makes it hybrid.

So what does a typical hybrid identity environment look like today? The most common hybrid environment is active directory, whether it's hosted on virtual appliances or virtual windows in whatever cloud infrastructure and intra-ID or ACTA. This is, again, part of that gain access to Office 365. And I'll often say to organizations, if you don't know if your hybrid, answer a couple questions. One of them is, do you have active directory on premise? Do you synchronize those identities to the cloud so that you can get access to Office or 65? If the answer to those two things is yes,

you're almost there. The third one is do the user sign in with one user principal name, one username? If that third one's yes, then you're hybrid. I know you describe these environments today as more of a mesh than being a simple synchronization path, what has prompted that change and why are organizations ending up with these multiple interconnected identity providers? It's typically the business processes that have driven it.

It's more of a mesh nowadays because there are more services that synchronize between organizational identities. So it's not just that I have Active Directory, for example, not just that I have Office 365, which means I have intra-ID and or Octa. It's also that I have other services where identities are created. Perhaps I have a provisioning service that synchronizes from some like workday or other identity provisioning service or HR service into the cloud, and thus down to Active Directory or directly to Active Directory and up to the cloud.

And so mesh is kind of a strong term because there's a line between everything. In reality, it's just kind of like a complex spider web, right? Okay. Well, as these become more interconnected, what kinds of challenges does that create for the defenders out there? Oh, so many. So the big challenge is, where does one identity begin,

and when does it end? And in many cases, it ends up being even more complex than that. It's not just when does it, where does it begin and where does it end in terms of updating and creation and, you know, maybe pruning those identities, but it also becomes what ingredients come from which identity stores. So like I could, I could say it's created an active directory, for example, but it gets some memberships and access and resource access and app access in ACTA or an intra. Or maybe it's created an HR service and, or maybe it's created an HR service and it gains all these things through several different identity providers.

That's where the complexity comes in, because without truly understanding where all those ingredients come in, putting them back is hard, and detecting those changes is hard. What is the status for most organizations in terms of having visibility into all of those tendrils? I think the larger organizations now have unaware of what identity stores they have.

For the most part, organizations understand the dependencies and have a rough idea of the flow. Understanding it holistically, start to finish, and being able to view it, secure it, and troubleshoot it if it goes wrong

between those, or heaven forbid, restore it after an incident, that's a place where it's, it's kind of almost a frontier. I know one of the themes in your research is this idea of resilience versus prevention, and perhaps resilience is becoming the more important goal for identity security.

Can you unpack that for us? Sure. Prevention doesn't go away. Prevention has been the key theme in security for as long as there's been security outside of even cybersecurity. I think the challenge is that resilience is becoming more important given the fact that all of these underpinnings, all this complexity is getting high. highlighted with the advent of agentic AI and the use of a gentic AI for threat actors and now for defenders. So it's kind of put an accent on how do I understand that, how do I deal with it, you know, in terms of hybrid identity and how do I put things back? It's not just can I secure it,

can I prevent, you know, things from getting changed by a threat actor or a malicious insider? It's more about, can I actually be resilient in the face of this problem, which we know will occur. What specifically are we seeing with agentic AI in terms of the challenges that it's presenting to the defenders out there? It's democratizing threat actors access, right? It used to be, you know, you'd have people sharing, you know, maybe they'd come together bad people wanting to do bad things. they would come together and they would share techniques, share tools, and mutually kind of gain expertise that way. They don't have to do that now.

With agentic AI, they can train an agent to do specific actions and do it repeatedly and do it well. It's like really, really awful, nasty interns who do a great job, right? And they can have a whole bunch of them. And so that's the part that is kind of changing things. And so when I say, listen, you can't really think about it as just security controls in putting the security controls back. You have to think about this holistically about putting a whole identity back. And that's hard.

And I think Agentic AI has a place on the defender's side for sure. And we're in early thinking on, for my part, on how we do that, you know, in a concerted way. Like with all the tools at your disposal, how is Agentic AI help you? Are there areas where you feel like the defenders still have an advantage? Yeah, I think they do. I think that if you start with insights and clarity on what is important to your organization, tools will help with that, some dedicated time sitting down and looking at those tools, looking at your environment, having the business conversations with everybody in the business.

If you're like a health care provider, patient intake is important. What's the patient intake app, right, in service? That's important for us to understand. How do we understand, you know, the patient info? We need to maintain access to that. So identifying what those things are and making sure that all the systems, the identity systems that support access to and authorization to those things are known. And you've documented how to restore the access

and restore the working order of those things, that's a critical step. That puts you, even with threat actors out there, that puts you a step ahead. Because if you identify what's important in your organization, the next small step is get the tools to be able to put them back. And then you're a step ahead. So you might have somebody interrupt your little journey,

but they're not going to bring you down. What about organizations that have complex hybrid environments? Are there common mistakes that you see organizations making when they're planning out their identity resilience? I think the biggest mistake is not understanding that it is, in fact, hybrid. It can seem kind of silly to be talking about this kind of 20 years and almost from hybrid becoming kind of prevalent. but it is actually kind of a big deal is understanding where the source of the source of truth for an identity begins and then further understanding what you can do without the source of authority. A lot of this is just understanding where are we granting controls and what's important to our organization?

It kind of boils down to that. And that clarity piece is the biggest issue. The next one I would say outside of gaining clarity about what's important to the business and having the tools to put it back, is being ready for an incident. That's the number one, I would say. Incidents are going to happen, accidental or otherwise. It's just going to happen. Being able to deal with the incident,

having all the documents that maybe you hopefully have comprised beforehand at your fingertips, knowing how to contact and having methods to contact, all the team. If your teams or Zoom or whatever is down, that's critical stuff. That's probably the sort of corollary place where you have to devote some energy.

No, it's a really interesting insight. It reminds me, I suspect many of my coworkers have no idea what a phone tree actually is, right? That's true. It can happen. Yeah. Where do you suppose we're headed here? Over the next few years,

how do you anticipate this is going to play out? What I think's going to happen is, agentic AI is going to, I said the word democratize, is going to make it really available for more advanced attacks, things that we've been kind of worried about happening. Like, yeah, everybody's like, I wonder if they do this. I wonder if they do that.

I think those harder things to accomplish in terms of breaching security and overcoming security controls, I think they're going to become more prevalent, particularly in the identity space. So I think that's one. part of it. I think the second part of it is the urgency of using agentic AI for defenders is going to become equally important.

And agentic AI is a tool that must have information. It must have specific guidance. It must have specific tools to accomplish its job. And having clarity of what's important to the organization to tell the agentic AI, what to defend, and having the tools to actually defend and put things back is what's going to be the added piece. So I think this space, this layer of agentic AI, it's almost like rock'em-sockom robots. If you remember that game? So here's two, you know, we have a defender and attacker, and we're all going to have our agentic AI is working together. But you still have to know what the right techniques are and what to protect.

That's Tim Springston from Sempros. And finally, Meta spent part of this year exploring just how AI native it could become, and how many humans might become optional along the way. Reuters reports that Project OT, short for organizational transformation, considered shrinking some teams by as much as 60%, while AI agents took over much of their daily work. One round of layoffs followed in May, but CEO Mark Zuckerberg reportedly cancelled a planned second round. The rethink came amid questions about whether the promised productivity gains were actually materializing. Internal data reportedly showed code changes soaring, while user-facing improvements rose much more modestly. More concerning, AI agents were linked internally to a 40% increase in major technical and security incidents, with employees spending considerably more time cleaning up afterward.

By July, Zuckerberg acknowledged that agentic development hadn't accelerated as expected. Apparently, becoming AI native still requires a fair number of native humans. 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 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. 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 Ibin. Peter Kilby is our publisher, and I'm Dave Bittner. Thanks for listening. We'll see you back here tomorrow.