Breach Roundup: CISA Says Agencies Should 'Patch Smarter'

Cybercrime , Fraud Management & Cybercrime , Incident & Breach Response Breach Roundup: CISA Says Agencies Should 'Patch Smarter' Also, France Probes Tchap Breach, M&S Cancels Bonuses, June Patch Tuesday Pooja Tikekar (@PoojaTikekar) • June 11, 2026     Credit Eligible Get Permission Image: Shutterstock/ISMG Every week, ISMG rounds up cybersecurity incidents and breaches around the world. This week, CISA imposed aggressive new patching deadlines for federal agencies. Software supply-chain hackers triggered artificial intelligence security scanners with fake nuclear weapons prompts. A suspected Russian nation-state hacker appeared in Boston federal court. Microsoft warned that attackers are hijacking the AI boom with fake ChatGPT and Copilot tools. The computing giant also patched 200 vulnerabilities, including six zero-days. French authorities investigated a breach of the government's Tchap messaging platform, Marks & Spencer cancelled bonuses after a costly cyberattack. CISA ordered emergency action on an actively exploited Check Point VPN flaw. Another NHS trust disclosed patient data theft tied to the Synnovis breach. A romance-themed espionage campaign targeted Russian troops, and ransomware group Qilin continued to make headlines on multiple fronts. See Also: Know Thy Enemy: Threats to Cyber Resilience CISA Gives Federal Agencies 3 Days to Fix High-Risk Flaws The lead U.S. federal civilian agency rejiggered how the government should tackle vulnerability remediation, unveiling a prioritization model that starts with identifying whether assets are publicly exposed to the internet. A Wednesday binding directive from the Cybersecurity and Infrastructure Security Agency gives agencies three days should a vulnerability hit a trifecta of indicators: the underlying asset is publicly exposed, CISA has listed the flaw in its Known Exploited Vulnerabilities Catalog and hackers could obtain total control - or in the case of a flaw whose exploitation could be automated, partial control. "Patch smarter, not harder," urged Chris Butera, acting executive assistant director for cybersecurity and Jonathan Spring, senior technical advisor. The directive adopts the Stakeholder-Specific Vulnerability Categorization framework, which seeks to match remediation efforts to real-world risk. "This is going to be something of a shock for many cybersecurity practitioners who were trained to believe that CVSS is the standard," Tod Beardsley, vice president of security research at runZero and former section chief for vulnerability response at CISA, wrote in a LinkedIn post. The directive also requires agencies to determine whether vulnerable systems may already have been compromised before remediation begins. Federal agencies have faced increasingly aggressive patching mandates in recent years as ransomware groups, nation-state hackers and other threat actors have shortened the time between vulnerability disclosure and active exploitation. "I remain dubious that a three day deadline spread across more than a hundred agencies is an achievable patch cadence today," Beardsley also said. Supply-Chain Hackers Trigger LLM Triage With Fake Nuclear Triggers The hackers behind the Mini Shai-Hulud npm, Miasma and Hades wave of supply-chain hacks targeting software repositories came up with a new technique for throwing off detection: Triggering any artificial intelligence scanners by prompting them with forbidden guides on building biological or nuclear weapons. Software supply-chain security firm Socket spotted the feint being used in attacks launched over the weekend against the PyPI repository of Python programming language packages. The triggering prompts appear in a JavaScript block comment and not within the executable section of the code. "It attempts to derail scanners or analyst copilots that feed the beginning of a file to a language model without clearly isolating the content as untrusted data," Socket wrote. Triggering a LLM into getting its hackles up over obviously forbidden prompts "cause refusal behavior, prompt confusion, context pollution, or premature classification before the scanner reaches the actual malware." The gambit isn't a magical bypass, Socket said. Methods such as YARA rules and string extraction still work. "But it is a practical anti-analysis trick against naive LLM-first triage systems." Hackers have unleashed supply-chain attacks on code repositories for months now in a groundswell of hacks that grew after progenitor TeamPCP open sourced its original self-propagating code (see: Flurry of Supply-Chain Software Library Attacks). Suspected Russian Intelligence Hacker Charged in Boston A Russian national suspected of hacking for Russian intelligence made his first appearance in a Boston federal court following extradition from Thailand. Federal prosecutors say Denis Obrezko, 36, formed part of a hacking unit tracked as Void Blizzard, whose existence became public in May 2025 by the Dutch government and Microsoft. The Netherlands said the group used tactics similar to Unit 26165 of the Russian Main Intelligence Directorate, commonly tracked as APT28 but nonetheless appeared to be a distinct threat actor (see: NATO Countries Targeted By New Russian Espionage Group). Thailand's Ministry of Foreign Affairs told Reuters that it extradited Obrezko in accordance with Thailand's domestic law and its obligations under the related treaties on extradition, "while fully respecting the due process of law of the defendant." An FBI affidavit partially unsealed in federal court says investigators were able to identify Obrezko as a Void Blizzard hacker through the classic technique of following a chain of online account and cryptocurrency payments used to set up hacking infrastructure until they discovered a Gmail address. That led them to finding the cell phone number used for backup - the same number used to register another account, denis.obrezko@gmail.com. Obrezko, the affidavit details, used the same email and cell phone number to register a PayPal and an X social media account. The PayPal account supplied investigators with Obrezko's birthday, May 17, 1990. The affidavit details Void Blizzard activity, which has included mass email harvesting from U.S. businesses. Dutch intelligence agencies in 2025 characterized the group as looking for information showing the procurement and production of military goods by Western governments and weapons deliveries to Ukraine. Stolen authentication credentials and password spraying made up the bulk of Void Blizzards operations, the Dutch government said. Threat Actors Exploit AI Brands to Deliver Malware, Steal Credentials Threat actors are weaponizing the popularity of AI platforms in social engineering campaigns by using fake AI tools, spoofed websites and malicious advertisements to distribute malware and harvest credentials, Microsoft warned Monday. The tech giant said campaigns observed in recent months impersonated AI services including ChatGPT, Microsoft Copilot, DeepSeek and Anthropic's Claude. The attacks ranged from phishing and malvertising campaigns to search engine optimization poisoning scams designed to trick users seeking AI tools into downloading malware or disclosing sensitive information. Microsoft attributed a malvertising operation active since at least early this year to an initial access broker it tracks as Storm-3075, which used fictitious AI product names such as "Awesome AI Windows Plugin" and "Flux Pro AI" as lures in malicious popups and fake installer filenames. The threat actor often delivered malware bearing a code-signing signature using certificates obtained through a malware-signing-as-a-service operation attributed to a threat actor Microsoft tracks as Fox Tempest. Microsoft Patches 200 Flaws, Six Zero-Days in June Update Microsoft on Tuesday released security updates for 200 vulnerabilities, including six zero-day flaws, as part of its June monthly dump of fixes. This Patch Tuesday includes fixes for 33 critical vulnerabilities spanning Windows, Microsoft Office, Azure, Visual Studio and .NET, with 28 of the critical flaws involving remote code execution. Three of the zero-days were publicly disclosed before patches became available: CVE-2026-45586, a privilege-escalation flaw in the Windows Collaborative Translation Framework that can grant an attacker SYSTEM privileges, CVE-2026-50507, a Windows BitLocker security feature bypass that could allow an attacker with physical access to circumvent full-disk encryption and CVE-2026-49160, a denial-of-service vulnerability in the Windows HTTP.sys stack affecting HTTP/2. Microsoft also addressed CVE-2026-42897, an Exchange Server vulnerability that had been under active exploitation since May and was previously mitigated with workarounds before receiving a full patch this month. The June release includes dozens of remote code execution flaws, elevation-of-privilege bugs and information disclosure vulnerabilities. Several affect core Windows components and enterprise environments. M&S Cancels Employee Bonuses After Cyberattack Contributes to Profit Drop British retailer Marks & Spencer suspended bonus payments, affecting 63,000 employees, board members and senior executives, including CEO Stuart Machin, after a cyberattack contributed to a 28.8% decline in annual profit and triggered a regulatory investigation (see: M&S Reportedly Hacked Using Third-Party Credentials). The department store chain said Wednesday that statutory pre-tax profit fell to 364.6 million pounds - $419.29 million - for the fiscal year ending March 31, down from 511.8 million pounds - $588.57 million - a year earlier. The company attributed 131.3 million pounds - $151 million - in costs to system recovery efforts, risk management measures and specialist advisory services following the cyber incident. The kibosh on bonuses will also affect boards members, British daily newspaper The Times reported. The group of adolescent extortionists known as Scattered Spider is widely suspected of being responsible for the incident, which security experts have said was a ransomware attack on the company's VMware ESXi server (see: Scattered Spider Linked to Marks & Spencer Hack). French Government Investigates Breach of Tchap Messaging Platform A compromised user account on Tchap, the encrypted messaging platform used by French government employees, exposed public conversations and user profile data tied to more than 73,000 agents. France's national cybersecurity agency ANSSI detected the intrusion on Sunday. The Interministerial Digital Directorate said the breach stemmed from account impersonation via a social engineering attack, with the malicious account since identified and blocked. The directorate said private chats on Tchap are end-to-end encrypted and cannot be accessed even through a compromised account. A threat actor claimed responsibility for the breach allegedly stole data tied to 73,467 user accounts, including names, email addresses, organizational affiliations, device metadata and avatars. The actor also claimed to have obtained more than 643,000 messages, 876 chat rooms, nearly 60,000 media files totaling 13.5GB and references to documents marked "Diffusion Restreinte," a French restricted-distribution classification. Access was reportedly gained through a single compromised account on Tchap's education shard. The French government-developed Tchap is part of a wider Gallic effort to favor local alternatives to American-made tech (see: France Tees Up Big Public Sector Move Away From US Tech). CISA Orders Emergency Patching of Check Point VPN Flaw Attackers linked to the Qilin ransomware operation exploited a previously unknown Check Point VPN vulnerability in attacks targeting dozens of organizations, prompting CISA to order federal agencies to secure affected systems within 72 hours. CISA this week added CVE-2026-50751 to its Known Exploited Vulnerabilities catalog after Check Point disclosed active exploitation of the flaw, which allows unauthenticated attackers to gain remote VPN access under specific configurations. Check Point said it first observed exploitation attempts on Sunday and recorded a sharp rise in attacks in recent weeks. The company linked at least one intrusion to a Qilin ransomware affiliate and said several dozen organizations worldwide have been affected. Russian-speaking Qilin is one of the most active ransomware-as-a-service groups in the world, although its affiliates aren't always the brightest of people (see: Breach Roundup: the Qilin Hack That Wasn't). Essex NHS Trust Confirms 2,380 Patient Records Stolen Mid and South Essex NHS Foundation Trust said 2,380 patient records were compromised in a June 2024 cyberattack against pathology services provider Synnovis, making it the latest British National Health organization to disclose the scale of the breach. The trust said that the stolen data originated from computer drives belonging to Synnovis, a third-party provider that processes blood, urine and tissue samples. The disclosure follows a similar announcement by Bedfordshire Hospitals NHS Foundation Trust, which said nearly 33,000 patient records were stolen in the same incident. Synnovis previously confirmed that the stolen information was published online. According to the company, the data may include patient names, dates of birth, NHS and patient identification numbers, postcodes and diagnostic test results. The Russian-speaking Qilin claimed responsibility for the attack, which disrupted pathology services across multiple NHS organizations. Romance-Themed Espionage Campaign Targets Russian Troops A new, undocumented cyberespionage group is targeting Russian military personnel by posing as women seeking romantic relationships on Telegram, said researchers at Russian cybersecurity firm F6. The group, dubbed SiribClone, has been active since at least mid-2025 and primarily targets troops stationed in border regions and combat zones. Researchers said the operation appears designed to gather battlefield intelligence by stealing files, monitoring communications and collecting sensitive military information. Attackers initiate conversations with servicemen on Telegram and other messaging platforms before luring them to download malicious apps or enter credentials on fake websites. One Android spyware strain, named SafeLoveStealer, can steal files, location data and other information from infected devices while enabling remote microphone access. The group also uses phishing sites masquerading as Telegram login pages to hijack accounts and intercept communications. Researchers identified a second malware family, SiribGrabber, designed to steal files from desktop systems. An internal management platform used by the group contained references to military ranks, unit designations and locations, reinforcing the assessment that the campaign is focused on military espionage. F6 did not attribute the operation to any known threat actor or country. Other Stories From This Week Claude Mythos 5 Can Build Exploits But Can't Power Campaigns Data Center OT Flaws Could Help Hackers Kill Power and AC Identity Scams Evolve Into Multi-Stage Attacks Senate Committee Leader Seeks Answers on NYC Health Hack Miasma Worm Hits Microsoft's AI Coding Ecosystem With reporting from Information Security Media Group's Anviksha More in Mumbai and David Perera in Northern Virginia.