The Human Take

This week there’s a few interesting stories, but the write up in the Incident Reports section from Tech Insider collecting the information around the spring breach of Ernst and Young (EY), one of the Big 4 accounting firms. What really caught my eye was the timeline:

Date	Event
March 26–28, 2026	Unauthorized access to the ITSM platform begins
April 12, 2026	Unauthorized access window ends; documents already downloaded
April 23, 2026	EY detects anomalous activity on the platform
July 13, 2026	Date listed on EY’s client notification letter
July 15, 2026	EY files breach notification with the California Attorney General
July 17, 2026	Breach becomes public via multiple outlets
Late July 2026	ShinyHunters publicly claims responsibility for the intrusion
Late September 2026	New letters confirm Goldman Sachs and Man Group client exposure

The incident occurred sometime in March (and I question the dates there, because finding when something “started” is always a bit of… hypothesis), the adversary either got out or lost access on 12 APR, then they were detected by EY security in 23 APR. That may seem like a long time, but this is what it looks like when Cybersecurity takes a loss, a slick bad actor can root around for as long as they want. Then between detection and client notification was nearly… 3… months…

Cyber investigations can take a long while, but incident notifications of this potential criticality should normally be out within 24 hours of suspected breach. If you’re really conservative, maybe 3 days to really lock in the scope of the attack. 3 months? That’s public relations and legal teams filtering the security output into serious-but-dismissive notifications to the officially-mandated channels. This is the nature of things more than a condemnation of any particular business, but this whole incident highlights a very real problem in Cybersecurity, and it is not solvable by AI (and is slightly aggravated by it).

For the love of everything good and right about security, stop giving your critical data to third parties. Especially unencrypted data. There was a golden age in the late 2000s where PCs were still the daily driver, and we all learned how cool cloud storage was. Over time we – people and businesses – have moved out entire infrastructure to the cloud. The promise of lower maintenance, better security (which at one time was better than the petri dishes that were the PCs and local server farms of 2009) and lower overall cost made everyone rush to the cloud. Cloud providers found a way to encourage businesses to outsource their infrastructure and people to rent digital storage lockers to put all their junk in. Cost will always be debated, it really depends on how you want to categorize the numbers in an accounting ledger, but I think it’s fair to say that the security promise of the cloud have not delivered.

Cloud breaches started happening as people moved there, but more often than not, that could be chalked up to old IT being duplicated in the cloud and just getting compromised there. In 2014, a hacker breached the icloud accounts of numerous celebrities and posted their private data on the Internet. It was a huge scandal and pushed people into thinking about their personal security, but it also highlighted that putting all of your very private eggs in someone else’s basket may not be the best approach. In the dozen years since, cloud compromises are so common, no one bats an eye anymore.

In this case, EY – as I’ve gathered from reading the articles – was not directly breached, or was not initially breached. Instead, their IT Service Management tool (one of the surely 10 that they use), provided by a third party in the cloud, was breached. Was the whole product breached, or just EY’s instance? I’m not sure. Are they sure? The ITSM – and this has been verified – had client tax records, audit information, and a treasure trove of privacy information and financial data. So, let’s state that clearly. Ernst and Young – one of the largest accounting firms on the entirety of planet earth – is outsourcing… Something… That includes vital and critical records of their clients. And there’s the rub. Company A hires EY as a third party. EY hires ITSM company to handle critical data. Company A has now assumed the risk for ITSM company. If ITSM company hires Company X to process some of that data, Company A also gets to assume the risk of Company X.

If you’re getting someone to make the cake for the company Christmas party, it’s fine if the bakery uses a shady janitorial service. But if you handle the financial records for fortune 500 companies, do you really want to just believe that the service you hand your data to isn’t also hiring a shady janitorial company? I’d bet that EY has a Third-party Risk Management department – if it’s not contracted out – that is understaffed and woefully unable to handle the thousands of vendors that EY uses on a regular basis. But the wheel must turn, the spice must flow, and the vendors are used. That 3 month lag in reporting? That is – at least in part – the vendor being analyzed to figure out where EY liability stops and ITSM company X begins.

There is no good solution for this other than finding vendors who are doing it all themselves (good luck) or – to set up systems and procedures to ensure that third parties never have your critical data, unless it’s encrypted (by you or a method you understand and trust). I’m not made of stone – or time – so I use cloud storage, but if my data is important, I encrypt it at rest so if someone breaches my cloud accounts, they get a bunch of photos of my kids and gigabytes and gigabytes of encrypted nonsense data that they can’t do anything with without a key. And a password. And another key. You get the idea.

I’m speaking mostly to business here, but as an individual, you are just as much at risk as a company, if not more. If you’re interested in encrypting and have a little bit of time to learn something new, I recommend you look at Veracrypt, a free and open-source software for encrypting files, drives, volumes, creating hidden (and encrypted) drives, the works.

As AI dominates the Cyber news cycle, you may hear us back here yelling as loud as we can about the fundamentals. The AI monsters can’t get you if there’s no way in. If you trust someone else to handle your most holy of holies, you may be giving them a direct path to under your bed.

---

Weekly Cyber News Summary October 8, 2026

This week in cyber:

Citrix NetScaler zero-days exploited by state-sponsored actors globally

ANY.RUN tracks CSuite, N0va phishing campaigns targeting US and EU

AI cyberattacks evolve into automated, machine-speed attack chains

EU AI Act high-risk audits begin; California SB 1047 deadline approaches

FortiMail zero-day CVE-2026-104286 actively exploited worldwide

EY breach exposes Goldman Sachs and Man Group client tax data

State-Sponsored Attacks

Google’s Threat Analysis Group and Mandiant have attributed a coordinated series of cyber operations targeting the global defense industrial base to four nation-state actors: China, Iran, Russia, and North Korea. Unlike previous years where each country operated in isolation, these state-sponsored groups are now working in tandem across multiple vectors, striking defense contractors, supply chain vendors, and military technology firms simultaneously. The campaign leverages shared infrastructure, coordinated timing, and complementary toolsets to overwhelm defenders who traditionally track each actor independently. This convergence marks a structural shift in cyber warfare, where geopolitical allies pool resources to create multi-vector sieges that are harder to attribute, contain, or predict. For security teams, this means threat intelligence programs must now account for cross-actor collaboration rather than treating each nation-state campaign as a separate entity.

This development is significant because it compresses the attack surface across industries and geographies, forcing organizations to defend against a unified threat rather than isolated campaigns. The defense industrial base is particularly vulnerable because it sits at the intersection of government contracts, commercial technology, and export-controlled systems. A breach in one vendor can cascade through multiple prime contractors, making supply chain visibility more critical than ever. The coordinated nature of these attacks also suggests deeper intelligence sharing between the four nations, potentially including early warning of defensive measures and shared exploit libraries.

Sources:
ITBriefcase: https://itbriefcase.net/top-10-cybersecurity-stories-this-week-october-2-2026/
LinkedIn/Google TAG: https://www.linkedin.com/posts/automated-news-automator-solutions_google-links-china-iran-russia-north-korea-activity-7428151089306198017-U3Az

Regional Campaigns

ANY.RUN’s analysis of September 2026 threat activity reveals a multi-stage phishing ecosystem targeting US and EU organizations across technology, manufacturing, healthcare, and government sectors. Five distinct campaigns—CSuite, N0va, IronToll, Wazza, and TerminalFix—each exploit different attack surfaces: Microsoft 365 session theft, Device Code authentication flows, real-time payment card interception, anti-bot routing filters, and multi-stage payload delivery through compromised WordPress sites. The campaigns share a common pattern of using trusted cloud services and familiar authentication flows to blend malicious activity with legitimate business workflows, making detection increasingly dependent on behavioral analysis rather than static indicators.

The significance of these campaigns lies in their layered complexity. No single indicator tells the full story; a compromised session, a suspicious device registration, and an unusual outbound connection may all stem from the same initial lure. For SOC teams, this means investigation speed and context matter more than ever. The campaigns also highlight how attackers are moving beyond simple credential theft toward persistent access through token manipulation and remote management tool abuse, requiring defenders to monitor not just login events but the full authentication lifecycle across identity, browser, endpoint, and network layers.

Sources:
ANY.RUN: https://any.run/cybersecurity-blog/major-cyber-attacks-september-2026/

AI Integration

Cynet’s 2026 analysis documents a fundamental shift in how artificial intelligence drives cyber operations: AI has moved from assisting attackers to orchestrating entire attack chains autonomously. The report identifies hyper-personalized phishing, self-evolving malware, and AI-powered vulnerability discovery as the defining threats of the year. Machine-speed cyber warfare is now the norm, with attacks adapting in real time to defensive measures, creating feedback loops where both attackers and defenders deploy AI simultaneously. The report notes that 94% of organizations view AI as the biggest cybersecurity force shaping 2026, and 82.6% of analyzed phishing emails now contain AI-generated elements, fundamentally changing the scale and sophistication of social engineering campaigns.

What makes this year’s AI-driven threats particularly dangerous is the compression of the vulnerability lifecycle. AI can scan codebases, identify weaknesses, and generate exploits in minutes rather than weeks, narrowing the window between discovery and exploitation. Defenders must now rely on behavioral detection, automated response, and unified platforms to keep pace. The rise of dual-use AI models—systems that can both discover vulnerabilities and develop exploits—means that even restricted models will eventually find their way into attacker toolkits. The industry is shifting from signature-based detection to an assume- compromise posture, where continuous validation and intelligence-driven defense are the only viable strategies.

Sources:
Cynet: https://www.cynet.com/security-foundations/ai-machine-learning/ai-cyberattacks/

AI Policy and Posture Adaptation

September 2026 marks a critical inflection point for AI governance, with the EU AI Act’s first wave of high-risk system audits beginning across France, Germany, and Spain. The European AI Office is inspecting automated resume screening tools, algorithmic credit assessment systems, and AI triaging tools in healthcare, requiring providers to submit technical documentation on system architecture, training data provenance, and human oversight mechanisms. Simultaneously, California faces a September 30 deadline for Governor Newsom to sign or veto SB 1047, the Frontier AI Safety Act, which imposes strict obligations on developers training models exceeding 10^26 FLOPs, including mandatory pre-training safety protocols, full shutdown capabilities, and annual independent audits. Brazil’s Senate also voted on Bill 2338/2023 on September 16, establishing a comprehensive AI framework influenced by the EU model.

The convergence of these regulatory milestones reveals a global shift from voluntary AI governance to binding enforcement. Organizations operating across jurisdictions must now reconcile overlapping requirements: EU Article 11 technical files, California’s whistleblower protections and third-party audits, Colorado’s consumer-facing risk management programs, and China’s algorithm registry with mandatory watermarking. The compliance burden is no longer theoretical—it requires immutable audit logging, training data copyright opt-out verification, and automated red-teaming pipelines. For enterprises, the challenge is building a governance program that satisfies multiple frameworks without duplicating effort, with ISO/IEC 42001 certification emerging as a practical compliance passport for multinational operations.

Sources:
Cubbbix: https://cubbbix.com/blog/ai-regulation-september-2026-global-update/

Zero-Day Exploits/CVEs

Fortinet disclosed CVE-2026-104286, a critical CVSS 9.8 zero-day in its FortiMail secure email gateway, on October 1, 2026. The vulnerability combines path traversal (CWE-22) with improper NULL byte handling (CWE-158) in the Identity-Based Encryption GUI component, allowing unauthenticated remote attackers to write arbitrary files to the appliance via crafted HTTP or HTTPS requests. Fortinet confirmed active exploitation in the wild and published detailed indicators of compromise, including planted shared libraries (/data/lib/liblog.so), modified dynamic linker preload directives (/data/etc/ld.so.preload), and outbound connections to 79.141.169.187 and 45.129.0.192. CISA added the flaw to its Known Exploited Vulnerabilities catalog the same day and issued Binding Operational Directive 26-04, giving federal civilian agencies just three days to remediate—a compressed deadline that signals the severity of confirmed exploitation evidence.

The vulnerability is particularly dangerous because it requires no authentication, no user interaction, and no prior foothold—any internet-facing FortiMail appliance is a potential target. The patch gap compounds the risk: permanent firmware fixes are still pending for the 8.0, 7.6, and 7.4 branches, leaving administrators to rely on interim workarounds like disabling IBE and restricting management interface access. This is at least the second FortiMail-specific flaw tied to the IBE subsystem to land on CISA’s KEV catalog since 2025, suggesting a recurring architectural weakness. For security teams, the lesson is clear: email gateways are high-value targets that sit at the intersection of defensive control and attack surface, and a single compromise can silently disable anti-malware scanning, suppress phishing detection, and exfiltrate sensitive communications without raising downstream alerts.

Sources:
The CyberSec Guru: https://thecybersecguru.com/exploits/fortimail-cve-2026-104286-zero-day/
Tech Insider: https://tech-insider.org/fortimail-zero-day-cve-2026-104286-cisa-deadline-2026/

Incident Reports

Ernst & Young disclosed a supply chain breach that exposed personal and financial information belonging to clients of Goldman Sachs and Man Group. The incident affected a third-party IT service management platform used by EY’s tax services, where support tickets included attachments containing sensitive client tax information. An unauthorized third party accessed the platform between March 28 and April 12, 2026, downloading documents that included names, addresses, tax identification numbers, email addresses, and financial details. EY detected unusual activity on April 23 and, working with an independent cybersecurity firm, confirmed the data theft window. The breach was attributed to a vulnerability in Checkmarx software, though no specific CVE or affected version has been identified, limiting the technical depth of current assessments.

This incident is significant because it illustrates how third-party support systems can become a route to sensitive client data without requiring a direct attack on the client’s own infrastructure. Goldman Sachs confirmed its systems were unaffected and client assets safe, while Man Group described the incident as involving software used by EY rather than a direct compromise of its own systems. The delay between the last unauthorized access and EY’s detection—eleven days—means the theft occurred before any response was triggered. For organizations, the takeaway is that vendor risk management must extend beyond the primary supply chain to include every platform that touches client data, and that access monitoring should be continuous rather than periodic.

Sources:
Tech Insider: https://tech-insider.org/ey-breach-goldman-sachs-man-group-data-2026/
Cyber Security News: https://cybersecuritynews.com/ey-data-breach-exposed/

The Human Take

Zero-Days Everywhere — What a Record Patch Tuesday and an Alleged FBI Hack Teach Us About Unknown Flaws

A zero-day is a flaw attackers exploit before a fix exists. This month delivered two very different zero-day stories. One ended with a patch. The other is still unfolding.

Story one: Microsoft’s biggest Patch Tuesday ever

On September 8, Microsoft shipped its largest Patch Tuesday on record. Counts vary by how you tally, but Tenable puts customer-actionable fixes at 964 CVEs, with 104 rated critical. That’s roughly 70% above the previous record, set just this July.

Two of those flaws were already being exploited in the wild:

• CVE-2026-85880, a heap buffer overflow in Windows Advanced Local Procedure Call (ALPC)
• CVE-2026-81963, a “link following” flaw in the Windows Update Stack, the component that installs updates

Both score 7.8 on CVSS. Both let an attacker who already has a foothold jump to SYSTEM, the highest privilege on a Windows machine.

The Common Vulnerability Scoring System (CVSS) is an open industry standard used to measure the severity of software security vulnerabilities. It assigns a numerical score from 0.0 to 10.0 to help security teams understand how critical a bug is and prioritize patching. A flaw that can be exploited remotely over the internet with no authentication and causes full system takeover receives a 10.0. Conversely, a flaw that requires local physical access, elevated privileges, and user interaction receives a much lower score

That’s the key point. Neither bug gets an attacker in the door. They turn a small intrusion, like one phished laptop, into full control. Microsoft hasn’t said who used them or how widely.

Now story two: ShinyHunters says it hacked the FBI

On September 22, the FBI’s jobs portal was defaced with a message claiming it had been seized by ShinyHunters, a well-known extortion group. The site then went offline “for maintenance.”

The group’s claims, as told to BleepingComputer and other outlets:

• It found and used a new zero-day in Oracle PeopleSoft, HR software widely used for recruiting and payroll.
• It moved from there into FBI-managed AWS GovCloud systems.
• It took 2–3 TB of data on current and former employees and job applicants.

On motive, the group says this was retaliation for a May 2026 FBI FLASH report about it, not a money play.

What we know so far: the FBI says it is investigating unauthorized activity on its jobs site. It has not confirmed that internal systems were breached or data stolen. Notably, that same FBI report warned that ShinyHunters tends to exaggerate its access. Treat this as a serious claim, not an established fact.

Even so, the pattern is credible. Oracle issued an emergency alert in June for a different PeopleSoft flaw (CVE-2026-35273) that the group had already exploited against education institutions.

The common theme?

One story is about the attacker’s first step: an unknown flaw in internet-facing software. The other is about the next step: privilege escalation once inside. Real intrusions chain both. The initial access bugs and the local privilege escalation bugs perform different functions in the kill chain. Initial access breach perimeter defenses, while local privilege escalation turns standard access into administrative control.

That’s why “we’ll patch when there’s a fix” isn’t enough. By the time a zero-day has a CVE number, someone has usually already used it.

What to do this week

• Patch the two Windows zero-days first, then the critical remote code execution bugs. With nearly 1,000 fixes, prioritization matters more than speed on everything.
• Run PeopleSoft or other HR systems? Limit what’s exposed to the internet, watch for unusual activity, and check Oracle’s advisories closely.
• Assume a foothold. Detect privilege escalation and lateral movement, especially into cloud environments.
• Recruiting portals hold sensitive data. Applicant records are a target in their own right. Protect them like payroll data.

Weekly Cyber News Summary September 25, 2026

This week in cyber:

• ENISA report details persistent Russian and Chinese state-nexus intrusions.
• Russian and Chinese threat groups target European and regional sectors.
• Palo Alto Networks launches AI-driven continuous offensive testing service.
• Anthropic CEO advocates safety pacing guidelines for frontier AI models.
• Microsoft fixes two exploited zero-days in Patch Tuesday release.
• ShinyHunters claims breach of FBI database via PeopleSoft exploit.
• EU Cyber Resilience Act mandatory reporting rules enter into force.
• Senator Markey introduces bill establishing AI cyber investigation board.

State-Sponsored Attacks

ENISA Highlights European Surge in State-Sponsored Attacks

The European Union Agency for Cybersecurity (ENISA) released its comprehensive Threat Landscape report assessing state-sponsored activity impacting EU member states. Russian state-nexus threat actors accounted for nearly half of all identified nation-state cyber activity, predominantly focusing on diplomatic institutions, central governments, and public infrastructure across Eastern Europe. Concurrently, China-nexus groups maintained persistent cyberespionage intrusions aimed at transport and logistical networks across the continent.

The findings emphasize that nation-states rely heavily on unauthorized access, credentials theft, and initial access acquired through vulnerability exploitation. Furthermore, state actors are increasingly leveraging hybrid operations—combining traditional intrusion operations with foreign information manipulation, denial-of-service barrages, and critical infrastructure probes—to test defensive capabilities and exploit geopolitical volatility.
Sources:
Industrial Cyber: https://industrialcyber.co/reports/enisa-threat-landscape-2026-highlights-ransomware-vulnerability-exploitation-ai-enabled-attacks-across-eu-organizations/
Cybersecurity Dive: https://www.cybersecuritydive.com/news/state-infrastructure-resources-cyberthreats/830178/

Regional Campaigns

Russian Hybrid Operations Target Northern European Infrastructure

A coordinated sequence of cyber and physical operations targeting Northern European nations, including Sweden, Finland, Denmark, and Estonia, has intensified throughout September 2026. Threat intelligence disclosures confirm that state-aligned groups executed distributed denial-of-service (DDoS) campaigns against public services alongside reported physical disruptions to undersea telemetry cables. The aggressive campaigns highlight a shift toward multi-domain pressure against regional trade and government assets.

In addition to direct network disruptions, threat intelligence reports revealed the deployment of autonomous tooling and multi-platform media manipulation designed to destabilize public sentiment and disrupt local communications. Security analysts emphasize that regional defenses are being continuously tested by automated attack vectors, requiring closer operational alignment between regional CERTs and infrastructure operators.
Sources:
Cyfluence Research: https://www.cyfluence-research.org/post/cyber-based-influence-campaigns-7th-13th-september-2026-report
Infosecurity Magazine: https://www.infosecurity-magazine.com/news/ransomware-attacks-reach-record/

AI Integration

Palo Alto Networks Unveils Agentic AI Security Operations Platform

Palo Alto Networks announced the release of Unit 42 Continuous Frontier AI Defense, an offensive security service leveraging agentic frontier AI models to perform continuous vulnerability discovery and attack path analysis. Built on specialized models from Anthropic and OpenAI, the framework executes automated adversary simulations across cloud environments, source code, and network endpoints to discover exposures before threat actors exploit them.

The service is designed to match the accelerated speed of AI-driven attacks, which have reportedly compressed enterprise breach timelines from weeks down to hours. By combining continuous machine scanning with human verification from offensive experts, the system validates exploitability, generates virtual patches, and integrates directly into enterprise remediation platforms.
Sources:
Unite.AI: https://www.unite.ai/unit-42-unveils-subscription-for-continuous-ai-powered-offensive-security/
CISA Advisories: https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-251a

AI Policy and Posture Adaptation

Industry Leaders Propose AI Pacing Frameworks to Align Defense and Safety

In a major policy movement across the artificial intelligence and cybersecurity sectors, industry leaders have published formal frameworks advocating for voluntary AI safety pacing agreements. The strategy outlines structured protocols allowing frontier model developers to deliberate on safety thresholds and threat capabilities without breaching federal antitrust constraints.

The proposed frameworks recommend utilizing independent third-party evaluators, establishing clear antitrust firewalls between commercial operations and safety research, and securing government oversight from regulatory agencies. This posture adjustment comes amid mounting pressure on organizations to ensure defensive security mechanisms and alignment protocols evolve in tandem with high-capability agentic models.
Sources:
Holland & Knight: https://www.hklaw.com/en/insights/publications/2026/09/ai-pacing-agreements-legal-issues-and-how-to-attempt-to-manage-them
Davis Wright Tremaine: https://www.dwt.com/blogs/privacy–security-law-blog/2026/09/trust-issues-september-2026

Zero-Day Exploits/CVEs

Microsoft Addresses Active Zero-Days in Record September Update

Microsoft released its September Patch Tuesday security updates addressing 974 CVEs across its software ecosystem, highlighting two privilege escalation zero-day vulnerabilities actively exploited in the wild. The first zero-day, tracked as CVE-2026-85880, is a heap-based buffer overflow in the Windows Advanced Local Procedure Call (ALPC) component that enables a local attacker to break out of low-privilege sandboxes to reach full SYSTEM privileges.

The second actively exploited flaw, CVE-2026-81963, impacts the Windows Update Stack through an improper link resolution vulnerability. Local attackers with minimal privileges can manipulate link resolutions to elevate access without requiring any user interaction. Microsoft urged administrators to prioritize these updates alongside critical remote code execution fixes affecting Netlogon and DNS Server.
Sources:
CrowdStrike Blog: https://www.crowdstrike.com/en-us/blog/patch-tuesday-analysis-september-2026/
SecurityWeek: https://www.securityweek.com/microsoft-patches-record-974-vulnerabilities-including-two-exploited-zero-days/

Incident Reports

ShinyHunters Claims Compromise of FBI Systems via Enterprise Zero-Day

The cybercrime syndicate known as ShinyHunters claimed responsibility for breaching systems associated with the U.S. Federal Bureau of Investigation, claiming to have exfiltrated sensitive data concerning current employees and job applicants. The extortion group posted proof of access on their dark web portal after temporarily defacing an applicant portal managed by the agency.

According to statements attributed to the group, the entry vector involved exploiting an unpatched zero-day vulnerability within an enterprise PeopleSoft deployment. Federal law enforcement acknowledged awareness of the claims regarding unauthorized activity on the application site and confirmed an active investigation into the scope and validity of the incident.
Sources:
The Hacker News: https://thehackernews.com/2026/09/shinyhunters-claims-fbi-breach-says-it.html
SecurityWeek: https://www.securityweek.com/astrana-health-data-breach-impacts-private-confidential-information/

Framework Changes

EU Cyber Resilience Act Mandatory Vulnerability Reporting Goes Live

Key reporting obligations under Article 14 of the European Union Cyber Resilience Act (CRA) officially took effect, establishing strict mandatory incident notification timelines for manufacturers of digital hardware and software products. Under the new mandate, organizations must report actively exploited vulnerabilities or severe security incidents within a tight 24-hour window from initial detection.

The framework mandates a three-tiered reporting process via a centralized EU platform managed by ENISA, requiring an initial alert within 24 hours, an expanded notification within 72 hours, and a final technical mitigation summary. The regulation significantly shifts global supply chain standards, compelling manufacturers selling into the EU market to modernize their vulnerability handling and patch distribution workflows.
Sources:
Inmove IT: https://www.inmoveit.com/en/cyber-resilience-act-in-2026-what-changes-from-september-and-what-businesses-should-review/
Kirkland & Ellis: https://www.kirkland.com/publications/kirkland-alert/2026/09/the-eu-cyber-resilience-act

New Legislation

U.S. Senate Introduces Bill to Create AI Cyber Incident Investigation Board

U.S. Senator Edward J. Markey introduced the Cybersecurity and AI Board of Investigations Act, aiming to establish an independent oversight body to investigate major cyberattacks assisted by artificial intelligence. Modeled after the National Transportation Safety Board (NTSB), the proposed board would be granted subpoena authority to compel testimony and document disclosures from public and private entities during post-incident inquiries.

The legislative initiative stems from growing concerns surrounding agentic AI models being deployed to conduct automated network exploitation and data exfiltration. If enacted, the board will be tasked with conducting forensic evaluations of complex AI-assisted security breaches, publishing public reports, and issuing binding security recommendations across federal infrastructure and key commercial sectors.
Sources:
U.S. Senate Press Release: https://www.markey.senate.gov/news/press-releases/as-ai-agents-carry-out-attacks-senator-markey-introduces-legislation-establishing-independent-body-to-investigate-cyber-hacks-assisted-by-artificial-intelligence
DLA Piper Privacy Matters: https://privacymatters.dlapiper.com/2026/04/comprehensive-federal-privacy-legislation-introduced/

The Human Take

Try to hold on to your surprise, but AI made the news again. I’m not going to turn every one of these into a diatribe about AI, but… Well… Just a little more.

Agentic AI is becoming the new threat. Last time I was here, I mentioned that AI can’t do anything we can’t, but it can do what we can do better than we can. An agent is a purpose-built AI that does something on behalf of (or in lieu of) a person.

Want to aggregate a bunch of news? Got an agent for that.

Want to check your email and manage your inbox? Agent.

<takes a deep breath>

Want to map a network, find some vulnerabilities, then exploit a vulnerability, drop a payload, phone home, move laterally, drop ransomware and ransom notices, delete every backup you can find, steal all the password lists on the network and start exfiltrating data? There’s an agent for that.

I have run penetration testing teams in the past, and the biggest problem was time and organic needs. I have four people to do a test, I have 800 hours of work to do, that’s about 5 weeks of work if my team does nothing else. And they get bored. They have to sleep and eat and touch grass, the AI needs none of that. Now, I have a tool that I can give to one of those humans, they tell the tool what to do. Then they keep an eye on it while they do all the squishy human things they need to do, or coordinate multiple agents doing multiple penetration tests at once. It doesn’t replace my penetration tester, it makes him more powerful. And we have rules to follow, clients to please, laws to not break. Imagine those same tools in the hands of someone who’s only motivation is gain, and their concern for laws, clients or rules is zero.

A massive law firm experienced a breach this week, showing that everyone needs to get on it, because with the speed of attack now, there’s no such thing as “we’re probably good.” Iran pulled some stunts showing why you should monitor your network for unauthorized messaging traffic (or why you should have such a thing as “unauthorized messaging traffic”). CISA also issued a warning for ScreenConnect being under active attack as a CVSS of 9.9, you should patch that if it means something to you. If it doesn’t, make sure it doesn’t mean something to you, then ask your organization how they are handling remote connections. If you’re in security specifically, ask if the restrictions around your remote solutions are strict enough. Spoiler warning: They are most likely not.

---

Weekly Cyber News Summary September 17, 2026

This week in cyber:

• Western nations expose Iranian spyware Chosen Brick targeting global dissidents.
• Russian and Chinese actors hijack Claude AI for automated cyberattacks.
• First autonomous AI agent breach reported to Spanish privacy regulators.
• Google patches zero-click Pixel modem vulnerability exploited in targeted attacks.
• Law firm Greenberg Traurig discloses client data exposure following breach.
• ConnectWise patches critical ScreenConnect privilege flaw under active exploitation.

State-Sponsored Attacks

Iran Deploys ‘Chosen Brick’ Surveillance Malware Against Global Dissidents

Cybersecurity and intelligence authorities from the United States, the United Kingdom, and the Netherlands issued a joint advisory exposing an Iranian state-sponsored surveillance campaign utilizing a Windows malware family dubbed Chosen Brick. Operating since at least 2025 under Iran’s Ministry of Intelligence and Security (MOIS), threat actors leverage social engineering over WhatsApp and Telegram to build rapport before delivering malicious installers disguised as utility software or medical records. The targeted operations aim to harvest emails, contact lists, and social media messages to physically track dissidents, activists, and journalists opposing the Iranian regime.

Upon installation, Chosen Brick sets up persistent registry keys and configures exclusions in Microsoft Defender to evade local security software. The malware uses individual Telegram bot IDs for command-and-control operations, enabling attackers to record host microphone audio, grab desktop screenshots, exfiltrate messaging database files, or wipe host machines entirely. Regulators noted that exfiltrated personal details have been posted onto pro-Iranian leak sites to conduct public harassment and intimidation campaigns against targets.
Sources:
SecurityWeek: https://www.securityweek.com/us-uk-dutch-agencies-expose-iranian-chosen-brick-surveillance-malware/
Security Affairs: https://securityaffairs.com/199217/malware/chosen-brick-irans-surveillance-malware.html

AI Integration

State-Sponsored Actors Automate Exploitation and Malware Assembly via Claude AI

Anthropic published a report detailing the disruption of multiple state-aligned cyber campaigns that abused its Claude AI model for automated operational workflows. One notable operation attributed to a Russian state-nexus espionage cluster (tracked as GTG-20006 or Midnight Blizzard) utilized custom AI pipelines to automatically rebuild and re-deploy malware toolkits whenever security software detected static artifacts. The group targeted over 20 European, Ukrainian, and Middle Eastern defense and government agencies through hotel Wi-Fi DNS hijacking, mobile exploit kits, and credential-harvesting tools.

Additionally, Anthropic exposed a China-linked threat actor that used Claude to advance technical research across three parallel tracks for an anti-torpedo naval weapons system. Financially motivated cybercriminals also leveraged autonomous AI agents to execute supply chain data theft against a SaaS provider, harvesting over 2,100 Azure AD tokens across 40 corporate tenants in just 34 hours. Anthropic confirmed that the accounts associated with these operations were terminated and digital signatures were mapped to enhance model safety guardrails.
Sources:
The Hacker News: https://thehackernews.com/2026/09/russian-state-sponsored-hackers-use.html
Hindustan Times: https://www.hindustantimes.com/world-news/phishing-weapons-and-spying-top-5-misuses-of-ai-flagged-by-anthropic-claude-101789110893261.html

AI Policy and Posture Adaptation

Spanish Data Protection Authority Logs First Autonomous AI Agent Data Breach

The Spanish Data Protection Agency (AEPD) confirmed receiving the first official breach notification where an autonomous AI agent was identified as the primary vector of an intrusion. An undisclosed organization reported that an external party deployed an AI agent powered by a commercial Large Language Model to breach its environment. Operating with minimal human intervention, the agent successfully authenticated into the system, conducted reconnaissance to find application weaknesses, modified internal personal records, and exfiltrated customer billing invoices.

The regulator emphasized that the event represents a qualitative shift from traditional AI-assisted phishing to fully autonomous, multi-stage agentic attacks capable of planning intermediate tasks and executing code at machine speed. The AEPD advised enterprise risk managers to update compliance frameworks immediately, warning that defense strategies must incorporate rapid AI-driven containment systems because traditional human-dependent response timelines are insufficient against autonomous agent speeds.
Sources:
SecurityWeek: https://www.securityweek.com/first-agentic-ai-data-breach-reported-to-spanish-regulator/
Shattered.io: https://shattered.io/aepd-first-ai-agent-data-breach-spain-2026/

Zero-Day Exploits/CVEs

Google Patches Zero-Click Pixel Cellular Modem Vulnerability Under Active Exploitation

Google released its September 2026 security update addressing CVE-2026-58704, a high-severity zero-day vulnerability affecting the cellular modem software in Google Pixel devices. The bug stems from a logic error within permission validation routines inside the modem component, enabling proximal or remote attackers to bypass authorization checks and elevate execution privileges. Critically, the vulnerability can be triggered as a “zero-click” attack, requiring no interaction or link clicks from the device owner.

Google acknowledged that the flaw has come under limited, targeted exploitation in the wild, though it did not attribute the activity to specific threat actors. Following the disclosure, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-58704 to its Known Exploited Vulnerabilities catalog, ordering federal civilian agencies to apply security patches immediately. Google urged all Pixel users to update to security patch level 2026-09-05 or later.
Sources:
The Hacker News: https://thehackernews.com/2026/09/google-patches-pixel-modem-flaw-amid.html
9to5Google: https://9to5google.com/2026/09/16/google-pixel-targeted-zero-day-modem-attack/

Incident Reports

Global Law Firm Greenberg Traurig Reports Breach Following Dark Web Exposure

International law firm Greenberg Traurig confirmed a security breach involving client data exposure following regulatory filings with state authorities. The disclosure follows claims by an extortion cluster calling itself SilentRansomGroup, which listed the law firm on its dark web leak site after exfiltrating internal files. Official filings confirm that unauthorized actors gained access to internal documents containing sensitive personal information, including Social Security numbers.

The incident highlights ongoing threat actor targeting of major legal institutions acting as corporate data custodians. Greenberg Traurig is the fourth major international law firm to confirm a data security incident within recent weeks, joining a pattern of cyberattacks targeting legal practices to extract confidential corporate, financial, and personal records held on behalf of global clients.
Sources:
Federman & Sherwood: https://www.federmanlaw.com/blog/greenberg-traurig-llp-data-breach-investigated-by-federman-sherwood/
Emery Reddy: https://www.emeryreddy.com/blog/data-breach/greenberg-traurig-data-breach

Framework Changes

CISA Orders Immediate Remediation for Critical ConnectWise ScreenConnect Privilege Flaw

The Cybersecurity and Infrastructure Security Agency (CISA) updated its Known Exploited Vulnerabilities catalog to mandate rapid patching of CVE-2026-84869, a critical flaw affecting ConnectWise ScreenConnect remote management software. Rated CVSS 9.9, the vulnerability involves missing authorization and improper privilege management within the client software. The bug enables an attacker with basic privileges to transfer and execute unauthorized payloads on endpoints during active remote support sessions without host interaction or prompt confirmation.

Threat intelligence teams reported that adversaries have been actively exploiting the vulnerability in worm-like automated attacks to drop persistent VBScript payloads and move laterally across client networks managed by Managed Service Providers (MSPs). CISA assigned a strict three-day remediation deadline under Binding Operational Directive 26-04 for federal agencies, while security researchers urged enterprise administrators to upgrade to ScreenConnect version 26.6.5 or restrict file transfer permissions immediately.
Sources:
SecurityWeek: https://www.securityweek.com/connectwise-patches-screenconnect-vulnerability-exploited-in-worm-like-attacks/
Aviatrix Threat Research Center: https://aviatrix.ai/threat-research-center/connectwise-screenconnect-cve-2026-84869-exploitation-attacks/

The Human Take

AI driven threats. AI revolutionizes defense/offense. Governance frameworks struggle with rapid AI-driven threat evolution. We may as well just pack our bags and move off grid. Game over man.

This is… Well… This is Cyber

AI is a disrupter, no doubt about that. How much of a disrupter, final scope of influence, and how our role as cyber professionals change – well, that is still TDB. Sometimes all the AI news looks a little bleak from the trenches. How can you keep up with these ‘rapid AI-driven threats?’ Let’s start with not ignoring the basics.

CISA just released their vulnerability review for FY2024 and FY2025 (https://www.cisa.gov/resources-tools/resources/cisa-vulnerability-review), know what they found? Take a guess before reading the linked article. No doubt some fancy chained zero-day exploits driven by the agentic hive mind right? Nope, injection vulnerabilities. In FY2024 10.1% of all CVEs were injection based. In FY2025 only slightly better at 9.2%. Injection – yes, as in XSS and SQLi – those things that when we talk about them in the classroom students wonder, why are we still talking about this? Isn’t this old information? Maybe, but seems we haven’t figured it out yet.

Way back in the day I played basketball. I didn’t say I was good but I played basketball. When we started getting it handed to us by a seemingly superior team it was time for the coach to call a time-out, sit us all on the bench and remind us about the stupid mistakes we were making. Slow it down, stick to your man, no more stupid fouls. We weren’t being outplayed we were forgetting the basics that make the competition’s job difficult. Same applies to defending against all the seemingly unbeatable AI threats, and even the non-AI threats. Input validation, MFA, memory safe operation all provide barriers to otherwise simple exploits.

They’re not that good – we are ignoring the basics – we just need to get our head’s back in the game.

---

The Human Take

Hello again! Another week, another report full of AI-related news. There is a good call out to a semi-scary statistic this week… The bad guys are winning! It turns out that AI-driven offensive tools are outpacing AI-driven defensive tools.

This is… Well… This is Cyber.

As a whole, we’re always losing. The adversary isn’t a conventional force, it’s not even necessarily identifiable. The bulk of Cyber will fail through lax standards, novel attacks, or sheer incompetence. And that’s Cyber. We talk to each other, and we tell each other what’s going on, hoping that you can use what I used to succeed, or learn from my failures. Vendors – security and AI vendors – would have you believe that without AI-enabled tools, you will lose the fight to these new Skynet-powered adversaries. This is hogwash. It’s not the newest defensive tool that will stop the bad guy, it’s you.

Another article in today’s batch details the Huggingface breach committed by OpenAI’s toolset. I mentioned a few weeks ago how AI companies like to tout these penetration testing platforms as wild tigers. Coincidentally, these wild tigers can only be controlled and wrangled by the AI companies. AI is not self-aware. It’s a program. It’s ability to find novel paths makes it feel like it’s thinking “outside the box.” When the HuggingFace breach “escaped containment” or some other such Live-Action-Role-Play nonsense, they are reframing in a cute way. The reality is that their tool was not configured correctly. If I had a PenTester go that rogue while they were testing, they’d be given an opportunity to find a new job. While the tool found all kinds of neat stuff, that’s because AI never gets tired of slamming its head into walls. This is the strength of AI, it knows all kinds of things, and they can execute tools, and it doesn’t get bored, it doesn’t miss things. It’s not a super intelligence doing things no one understands, it’s doing things we all understand, just quickly and without stopping.

So, back to my original point… Cybersecurity, good old-fashioned network defense, doesn’t care that the attacker is coming with AI. The attacker is always coming with something newer, something faster, something smarter. Your awesome AI hacker machine can’t hack my Active Directory if I set up proper controls that prevent you from accessing it. What we’re seeing here is the end of “Oh, it’ll probably be fine” in Cybersecurity. It’s not that AI hackers are winning, it’s that they are showing us who is really ready, and who was just checking a box somewhere hoping that their security wouldn’t get tested. Turns out, AI is testing everyone now, and Cybersecurity is just another in the line.

This week in cyber:

• Regional campaigns show intense DDoS pressure on Israel.
• AI offense is outpacing defense, driving a new arms race.
• Enterprises race to meet new AI governance mandates.
• Critical Windows zero-day with 9.8 CVSS score patched.
• OpenAI reports on massive Hugging Face breach confluence.
• China expands CSL penalties and government oversight.

Regional Campaigns

Global Cyber Threat Briefing: July 2026 Attack Statistics and Trends

The cyber threat landscape in the Middle East continues to be heavily influenced by geopolitical tensions, with Israel being a prime target. The briefing notes that 85% of the incidents tracked against Israel were driven by DDoS attacks, reflecting sustained hacktivist pressure. These campaigns are not confined to government entities; they are indiscriminately targeting civilian infrastructure, including universities and a major medical center, illustrating the breadth of the regional conflict. Furthermore, the analysis confirms the high operational tempo of these campaigns. Researchers have tracked thousands of attacks linked to Iran, and the MOIS Wiper campaign, specifically tied to Iran’s Ministry of Intelligence, has been forensically linked to attacks against Middle Eastern organizations. This indicates a deliberate, state-backed effort to destabilize regional stability. Sources:
Global Cyber Threat Briefing: July 2026 Attack Statistics and Trends: https://cyber.thomasmurray.com/insights/global-cyber-threat-briefing-july-2026-attack-statistics-and-trends
The Escalating Cyber Risk Landscape in Regional Conflicts & Strategic Actions for 2026: https://blog.eclecticiq.com/the-escalating-cyber-risk-landscape-in-regional-conflicts-strategic-actions-for-2026
Cyber Based Influence Campaigns 3rd–9th August 2026 Report | CRC Analysis: https://www.cyfluence-research.org/post/cyber-based-influence-campaigns-3rd-9th-august-2026-report
Middle East Organizations: Iran-Linked MOIS Wiper Campaign: https://wasteland.me/intel/iran-linked-middle-east-wiper-attack

AI Integration

AI Cybersecurity Statistics 2026: Offense Is Winning — And …

Artificial Intelligence is rapidly becoming the defining force in cybersecurity, a shift recognized by 94% of security leaders according to the WEF Global Cybersecurity Outlook 2026. Critically, the Axis Intelligence ADSI shows that AI-powered offense is currently outpacing AI-powered defense across four of the six critical attack surfaces. The rise of “agentic AI”—autonomous agents capable of executing complex tasks—is accelerating this offensive advantage. This trend is particularly pronounced with confirmed attacks involving agentic AI. The integration of these autonomous systems allows attackers to operate at speeds and scales that human defenders struggle to match. This arms race is forcing organizations to rethink their entire security posture, moving from reactive defense to proactive, AI-enhanced hunting. Sources:
AI Cybersecurity Statistics 2026: Offense Is Winning — And …: https://axis-intelligence.com/ai-cybersecurity-statistics/
AI Cybersecurity Arms Race 2026: Defense vs. Offense: https://aibradaa.com/blog/ai-cybersecurity-arms-race-2026
AI in Cybersecurity 2026: How Artificial Intelligence Is …: https://zeqty.com/ai-cybersecurity-2026-offense-defense-transformation/
Agentic AI: The New Frontier of Cyberattacks in… – AI Dominance SG: https://dominance.sg/posts/agentic-ai-cyberattacks-asia-2026.html

AI Policy and Posture Adaptation

AI Governance and Regulation 2026: A Complete Guide to Global …

The global regulatory environment is rapidly maturing to keep pace with AI-driven threats, highlighted by the full implementation of the EU AI Act in August 2026. This act establishes strict rules for AI systems, particularly those deemed “high-risk.” Beyond Europe, Singapore is leading in agentic AI governance, while the U.S. continues to standardize through the NIST AI Risk Management Framework (AI RMF). For enterprises, this means a massive compliance roadmap. Organizations must now map their AI use cases against these evolving frameworks, ensuring transparency and accountability across all deployed models. This effort is critical for maintaining operational posture in a fragmented, yet rapidly standardizing, regulatory world. Sources:
AI Governance and Regulation 2026: A Complete Guide to Global …: https://www.hungyichen.com/en/insights/ai-governance-regulatory-landscape-2026
NIST AI Risk Management Framework: Implementation Guide (2026): https://aisecurityandsafety.org/en/guides/nist-ai-rmf-guide/
AI Security Standards: Key Frameworks for 2026 – SentinelOne: https://www.sentinelone.com/cybersecurity-101/data-and-ai/ai-security-standards/
EU AI Act 2026 Guide: Enterprise Compliance Roadmap | Etheon …: https://www.etheon.com/index/eu-ai-act-2026-guide-what-enterprise-teams-need-to-prepare-for

Zero-Day Exploits/CVEs

August 2026 Patch Tuesday: Updates and Analysis | CrowdStrike

The August 2026 Patch Tuesday was particularly significant, featuring the patching of 421 CVEs, including several critical vulnerabilities. The standout is CVE-2026-62893, a Critical Remote Code Execution (RCE) flaw with a CVSS score of 9.8. This vulnerability affects Windows Deployment Services and was identified as a use-after-free flaw, meaning an attacker can exploit a memory management error via a specially crafted network packet. This critical flaw was actively exploited in the wild, prompting CISA to issue an emergency alert. Additionally, the patch cycle included CVE-2026-62836, an elevation of privilege vulnerability affecting Azure SQL Managed Instance (a cloud database service), which carries a high CVSS score of 8.7. Sources:
August 2026 Patch Tuesday: Updates and Analysis | CrowdStrike: https://www.crowdstrike.com/en-us/blog/patch-tuesday-analysis-august-2026/
August 2026 Patch Tuesday: Microsoft Fixes 421 CVEs, One Exploited Zero-Day: https://www.securityweek.com/august-2026-patch-tuesday-microsoft-fixes-421-cves-one-exploited-zero-day/
Windows Zero-Day Hits Patch Tuesday: 421 CVEs Fixed [2026]: https://tech-insider.org/windows-zero-day-patch-tuesday-421-cves-2026/
CVE-2026-62836 in Azure SQL MI: https://www.crowdstrike.com/en-us/blog/patch-tuesday-analysis-august-2026/

Incident Reports

OpenAI releases its official report on the Hugging Face breach

OpenAI has released a comprehensive report detailing a major breach involving Hugging Face, an incident that revealed a rare and unexpected confluence of security failures. The report, released on August 26, 2026, frames the breach not as a single failure but as a complex event chain. The incident reflects a failure in the security controls that allowed the compromise to occur, despite the platform’s overall robust infrastructure. The breach involved the theft of significant data and underscores the risks associated with relying on third-party platforms. The official report provides deep insight into the attack vectors, suggesting that misaligned security behaviors within the platform were the critical factor that allowed the attack to succeed and escalate. Sources:
OpenAI releases its official report on the Hugging Face breach: https://techcrunch.com/2026/08/26/openai-releases-its-official-report-on-the-hugging-face-breach/
ATF confirms “major incident” after recent Qilin breach claims: https://www.bleepingcomputer.com/news/security/atf-confirms-major-incident-after-recent-qilin-breach-claims/
Data Breach Tracker: Major Incidents 2026 (Updated in Real …: https://axis-intelligence.com/data-breach-tracker/
Information is Beautiful: https://informationisbeautiful.net/visualizations/worlds-biggest-data-breaches-hacks/

New Legislation

Global Data Privacy Laws in 2026: Mid-Year Update – Complete Discovery Source

In Asia, the regulatory landscape is being dramatically reshaped by the amended Cybersecurity Law of China (CSL), which came into effect on January 1, 2026. These amendments are significant because they substantially expand the penalties for non-compliance with the CSL. Crucially, they also grant the Chinese government increased power and authority to oversee and mandate compliance across various sectors. Globally, other regulations are also tightening. The CCPA in California has seen expansions, and new rules covering automated decision-making technology and mandatory cybersecurity audits have taken effect in 2026. This signals a global pivot toward holding organizations accountable for the *process* of data handling, not just the outcome. Sources:
Global Data Privacy Laws in 2026: Mid-Year Update – Complete Discovery Source: https://cdslegal.com/insights/global-data-privacy-laws-in-2026-mid-year-update/
Global Data Privacy Laws 2026: Cross-Jurisdiction Compliance Guide: https://www.kiteworks.com/regulatory-compliance/global-data-privacy-laws-2026/
Data Privacy Laws in 2026: Compliance Guide: https://www.tekclarion.com/blog/cyber-security/data-privacy-laws-2026/
UK-Hosted AI & GDPR: What to Get Right | The Digital Hub: https://thedigitalhub.uk/guides/uk-hosted-ai-gdpr