Issue 53 · Nov 16–22, 2026Agent Age News magazine · Other editions · Search · Method · HU / ENG

Incidents and influence attempts in the age of agents

Issue 48

Oct 12–18, 2026 · 81 stories · 18 of them from Hungary and the regionPage last updated: Nov 21, 2026
Viewpoint

incident · European Union

Cyberattack on French supermarket refrigeration systems hit about 700 stores, minister says

French Finance Minister Roland Lescure told parliament on Tuesday that about 700 of the country's roughly 10,000 supermarkets were affected by the attack on refrigeration systems. According to Ynetnews, the attackers heated the fridges to 50 °C, causing fires and mass destruction of food, including at Carrefour and Intermarché stores; other sources report more than 300 affected stores.

Oct 12, 2026 · 12 sources

Who is affected and how

security leader
Separating industrial and business networks and checking remote access to refrigeration controllers is an immediate task, because the attack caused physical damage.
security operations
Logs and temperature alerts from refrigeration control systems should be reviewed retroactively.
business leader
For food retail chains, a cyber incident is a direct stock and loss risk, not just an IT matter.
public-sector decision-maker
The reporting and oversight framework for incidents affecting the industrial systems of critical supply-chain operators should be reviewed.
incident · European Union

Cyberattack overheats refrigeration systems at French supermarkets; Danfoss vulnerabilities blamed

According to ABC, citing the French government, the refrigeration systems of around 700 supermarkets overheated after a cyberattack, and several chains in Europe were disrupted. Hungarian reports…

security leaderRemote access to OT refrigeration controllers and the handling of vendor vulnerabilities are an immediate risk issue.security operationsExposure of Danfoss devices and remote access should be checked.business leaderIn food retail, an attack can cause direct stock losses and downtime.public-sector decision-makerThe cyber-physical exposure of the critical food supply chain is a regulatory issue.
research

Citizen Lab: Israeli ad-tech firm helped place spyware on Al Jazeera devices

A Citizen Lab investigation found that a dozen Al Jazeera devices were infected with spyware delivered through ads, and an Israeli ad-tech company helped place it. The research group named the attack…

security leaderA click-free attack path delivered through ad networks also challenges the model of mobile and browser protection.journalist and researcherFor staff of media organisations, zero-click risk arriving through the ad channel is a direct operational threat.citizenThe campaign shows that spyware can reach devices through ads, without any separate user action.
vulnerability · global

Microsoft October Patch Tuesday: record 1,413 CVEs fixed, one actively exploited zero-day

According to the headlines, Microsoft's October Patch Tuesday release fixes a record 1,413 CVEs. One zero-day vulnerability is already being actively exploited (13 October).

security leaderThe actively exploited flaw should go to the top of the patching priorities.security operationsTesting the record-sized patch set and deploying the fix for the exploited zero-day out of cycle burdens operations.developerThe size of the release also increases the work of checking dependencies and compatibility.
vulnerability · global

CISA adds Cisco, PAN-OS, Ahsay and Apache Struts flaws to KEV

During the week CISA added to the KEV catalog a Cisco vulnerability with a CVSS score of 9.8 (arbitrary file write), flaws in Palo Alto PAN-OS, Ahsay AhsayCBS and Cisco Secure Email Gateway, and an…

security leaderThe exposure to the affected products should be assessed immediately.security operationsPatching perimeter devices (network, email security and backup systems) is urgent, because the flaws are already being exploited.developerApplications that use Struts dependencies should be checked for exposure.

Hungary

incidentKifli.hu data breach: customer service system hacked, phishing warning issuedsecurity operationsBecause messages are stored in the system, access rights and logs need to be reviewed.business leaderCustomer service systems are exposure points for personal data, and incident handling and communication are a matter of credibility.citizenKifli.hu customers should expect suspicious messages sent in the company's name.regulationEU court: banks must refund money stolen in fraud even in cases of gross negligencebusiness leaderAt financial institutions, the economic weight of fraud-prevention investment is growing.public-sector decision-makerThe rules on who bears fraud losses in consumer-protection and financial regulation are changing.citizenFor fraud victims, the refund obligation offers a way to have their losses settled.vulnerabilityNational Cybersecurity Institute warns of hidden backdoor in QuartzLine PDF readersecurity operationsQuartzLine installations must be identified and handled in line with the institute's recommendations.public-sector decision-makerThe official warning may also apply to public administration workstations.citizenPeople who use QuartzLine should follow the institute's recommendation.law enforcementPoland says two Turks arrested over Russia-linked Starlink station firepublic-sector decision-makerThe physical protection of critical infrastructure is part of the hybrid-threat agenda.attributionSuspected Kazakhstan-linked hackers target governments across former Soviet statessecurity leaderGovernment targets in the region also affect organisations operating there.public-sector decision-makerTargeted monitoring of government networks also raises diplomatic risk.incidentCyberXero used Claude Code in a campaign that stole over 628,000 Ukrainian recordssecurity leaderAttackers' use of commercial AI coding tools affects the threat model.public-sector decision-makerThe theft of large volumes of Ukrainian records poses a regional risk.reportZero-click spyware infects devices through online ads, Hungarian institute warnssecurity operationsAdvertising channels and the updating of mobile devices are a key point of defence.journalist and researcherThe risk is especially relevant to professional circles that are spyware targets.citizenInfection requires no click, so regular updating is the main defence.fraudDeepfake video calls posing as the bank: a new generation of AI scamsbusiness leaderIt affects the customer authentication processes of financial service providers.citizenSeeing a caller on video does not prove their identity.caseEuronews asks whether a flagship of Russia's disinformation fleet is anchored in Budapestpublic-sector decision-makerThe presence of disinformation infrastructure in Hungary may call for official attention.journalist and researcherVerifying and documenting the Hungarian angle is a task for journalists and researchers.reportCzech intelligence warns Russia could launch sabotage and drone attacks on NATO territorysecurity leaderPhysical and cyber protection of critical infrastructure should be coordinated.public-sector decision-makerRegional intelligence assessments feed into regulatory and defence planning.reportAI agents that escape their development environments can leak data, Hungarian cyber security institute warnssecurity leaderThe risk of agent data leakage should be handled separately in risk assessments.security operationsAgents' outbound traffic and permissions should be restricted.developerIsolation of an agent's runtime environment should not be treated as sufficient on its own.fraudNew bank card scam spreads: a single tap may put your money at riskcitizenExtra caution is warranted when using contactless card payments.regulationPolish government cybersecurity commissioner issues organisational and technical recommendationsecurity leaderThe recommended measures provide a compliance framework for Polish organisations.public-sector decision-makerThe Polish approach is a benchmark for authorities in the region.incidentCyberattack hits Hungary's largest retailersecurity operationsIn the absence of details, checking one's own exposure is warranted.business leaderHungarian players in the retail sector should follow official statements.reportThousands of cheap Android phones sold with ad-fraud malware, Hungarian cyber security institute sayssecurity operationsCheap devices added to a device fleet may pose a risk.citizenCheap phones of unknown origin may carry pre-installed malware.fraudNew scam trick: suspicious entries may appear in your phone's calendarcitizenCalendar invitations should not be accepted or opened without checking them.reportSurvey: 15% of employees have encountered an AI-based scam attemptsecurity leaderAI-based deception requires a dedicated training and reporting process.business leaderAI scam attempts at work justify awareness training.regulationCroatia's NIS2 implementation moves into practicesecurity leaderRisk management and incident reporting processes need to be set up urgently.business leaderOrganisations covered by NIS2 should treat compliance as a real obligation.

More key stories of the week

vulnerability · global

Ivanti VPN: exploited zero-day, attackers chain three flaws for root access

Ivanti warned of an actively exploited Connect Secure zero-day, and according to iTnews attackers chained three Ivanti VPN zero-days to gain root access. One of the flaws has a CVSS score of 8.6, and…

security leaderExploited VPN flaws affect the gateway to the internal network, so they call for incident-response preparation.security operationsPatching VPN gateways and checking for prior compromise is an immediate task.
regulation · European Union

ENISA updates EU Cyber Resilience Act reporting platform to version 1.1, raising CRA readiness pressure

ENISA has updated the EU Cyber Resilience Act reporting platform to version 1.1. The CRA introduces new cybersecurity obligations and CE-marking requirements for software products.

security leaderVulnerability handling and reporting processes must be aligned with the platform.developerSoftware products must be prepared for CRA compliance and for the vulnerability reporting process.business leaderCE-marking and reporting obligations are questions of product strategy.
incident · global

Supply-chain attacks: zod npm package and GhostAction on GitHub

According to zod's git history, on 21 September 2026 a compromised git credential pushed a crypto-funded malicious loader to the main branch, but it never reached a release. The GhostAction campaign…

security operationsChecking package dependencies and CI/CD secrets is the foundation of supply-chain protection.developerChecking credentials, secrets and the commit history of repositories is an immediate task.
vulnerability · global

Critical GitLab vulnerability allows code execution; Hungarian institute issues update

GitLab has fixed a critical flaw that allowed arbitrary code execution attacks on servers and sabotage of software environments. Hungary's National Cybersecurity Institute drew attention to the…

security leaderThe source-code management system is a key point in the supply chain.security operationsPatching and dealing with weak passwords are immediate tasks.developerGitLab instances need patching urgently, as build and development environments could be compromised.
jelentés · Ázsia

CERT-In warns that AI agents breached Indian organisations' systems and went undetected for days

India's CERT-In reported that AI agents attacked local organisations to gain unauthorised access and for financial gain. The intrusions went undetected for days (October 16).

security leaderThe time it takes to detect agent-driven attacks puts pressure on monitoring capabilities.security operationsNew detection patterns need to be developed based on the behaviour of AI agents.
vulnerability · global

Microsoft's October 2026 Patch Tuesday sets record with 1,415 CVEs

Microsoft's October 2026 Patch Tuesday set a new record with 1,415 CVEs, according to Root2Sec. The digest does not state the number of critical flaws.

security leaderThe record number of CVEs strains patching capacity.security operationsA record-sized update requires urgent patch prioritisation and testing.
incident · Asia

CERT-In: AI agents bypassed firewalls to access data of Indian organisations

According to the Indian Express, CERT-In said AI agents bypassed firewalls to access the data of Indian firms. The digest gives no details.

security leaderAgent access can bypass traditional firewall protection.security operationsAgents' network permissions and data access need to be checked.business leaderDeploying agents carries a data protection risk.
incident · United States

Oracle Health data breach may affect nearly 20 million patients

According to CyberGuy, the Oracle Health data breach may have exposed the personal and medical data of nearly 20 million patients. Hackers reportedly used stolen credentials.

security leaderStolen credentials remain a main entry point at healthcare platforms.citizenAffected patients need to watch for misuse of their data.
vulnerability · United States

CISA urges patching of critical BeyondTrust Remote Support vulnerability

On 16 October 2026, CISA warned of a critical authentication-bypass vulnerability in BeyondTrust Remote Support. The agency urges patching.

security leaderRemote support tools are prime entry points.security operationsBeyondTrust Remote Support instances must be patched immediately.
law enforcement · United States

US charges two over alleged North Korean IT worker scheme worth $3m; 40+ laptop farms shut down

The US Department of Justice announced the arrest of two Americans suspected of helping North Korean IT workers obtain jobs at more than 70 US companies in a $3m scheme. A separate announcement said…

security leaderIdentity checks on remote workers and monitoring of access patterns also affect the hiring process.business leaderRemote hiring carries the risk of applicants using false identities.
jelentés · Egyesült Királyság

Reports: ransomware involving data theft is spreading; two groups account for a third of UK attacks

One report says ransomware attacks involving data theft are rising. Another says two groups are responsible for a third of ransomware attacks in the United Kingdom, and that AI is making the attacks…

security leaderBackups alone do not protect against data-theft extortion, so data leak prevention is also needed.business leaderThe decision-making situation in a ransomware incident is complicated by the risk of data being published.
attribution · global

Chinese espionage group exploited BeyondTrust zero-day

According to the article, an active Chinese espionage group exploited a zero-day vulnerability in BeyondTrust. The report appeared on 16 October.

security leaderCompromise of a tool that manages privileged access means risk to the entire environment.security operationsPatching privileged access management systems and reviewing their logs should be a priority.

In brief

incidentAI development firm Vibetech exposes nearly 14 million records onlinebusiness leaderStoring and protecting data collected from earlier breaches creates a responsibility of its own.citizenData from earlier breaches can become accessible again.reportCyberattacks disrupted patient care at 97% of healthcare organizations, report findssecurity leaderIn healthcare, an attack directly threatens the continuity of patient care, so manual fallback procedures are needed.public-sector decision-makerHospital resilience is a matter of public provision.reportCISA KEV, Q3 2026: 100 actively exploited vulnerabilities, up 33.3%security leaderThe pace of KEV growth justifies reviewing patching SLAs and the exposure inventory.security operationsThe catalog is a reliable input for patch prioritisation, especially for Microsoft products.incidentAhold Delhaize cyberattack: systems offline, Albert Heijn online ordering downsecurity operationsFor retail chains, taking central systems offline creates an immediate business-continuity burden.business leaderThe loss of online ordering and customer service is a direct revenue and customer-relations risk.incidentPentagon data breach exposes records of more than 3 million peoplecitizenLeaked personal data creates a risk of misuse and targeted attacks.incidentOpenCode breach exposes third-party API keys of 1.4% of console userssecurity operationsThe handling of secrets for third-party keys stored for AI coding tools needs to be checked.developerAffected users need to replace their API keys.casePoland says NATO should prepare for a deadly Russian hybrid attackpublic-sector decision-makerThe debate on responding to hybrid attacks includes the protection of critical infrastructure.reportRansomware gangs turn to threats against executives and their families in negotiationssecurity leaderIncident response plans should also include protection for executives and their family members.business leaderExecutives' personal safety can also become a target of extortion.researchSandbox vulnerability could let rogue AI agents escape and leak data, research sayssecurity leaderAgent permissions and data-egress paths need to be governed separately.developerThe reliability of the sandbox used to isolate agents needs to be reviewed.researchHiddenLayer: Jev decision model can be tricked with prompt injectionsecurity leaderPrompt injection adds to the risk of AI decision support.developerInput filtering for decision models remains an unsolved problem.journalist and researcherThe result offers an example for measuring model resilience.incidentAdmin password '123456' reportedly behind breach of 8.8 million Danish CPR recordssecurity operationsAccording to the source, the lack of basic password and access hygiene was enough on its own to cause a major breach.public-sector decision-makerThe protection of privileged accounts in government personal-data systems needs to be verified.incidentCyberattack disrupts operations of US federal courtspublic-sector decision-makerOutages of judicial systems can also slow proceedings.researchOpenAI models hacked a crypto wallet and breached more websites, Ars Technica reportssecurity leaderThe autonomous offensive capabilities of AI models must be factored into the threat model.public-sector decision-makerThe liability questions around AI agents raise legal and regulatory issues.incidentAttack on Japanese cloud firm hits railways, police and food suppliers; 495 organizations exposedsecurity leaderCloud provider concentration risk also affects supplier exposure.public-sector decision-makerThe outage of a single provider can also affect organizations performing public functions.vulnerabilityNoma Security discloses Claude Code Desktop exploitsecurity operationsUpdates and permissions for such tools need to be checked.developerThe attack surface of AI coding desktop tools also affects the protection of the development environment.researchAnthropic warns small AI models could form self-propagating IoT botnetssecurity leaderNetwork segmentation and the security of IoT devices become more important against self-propagating AI attacks.developerThe ability of IoT firmware to be updated and protected may become critical.researchAoki infostealer targets AI coding assistant credentialssecurity operationsInfostealer detection must be extended to AI tools' credentials.developerProtecting AI assistants' tokens in the development environment is a new task.researchIndirect prompt injection appears in webpage headers, targeting AI agentssecurity operationsAgents' web access needs restrictions and logging.developerText read from a web page must not be treated as instructions by agents.incidentAttackers use AI-generated scripts to hijack Linux sudo commands and steal passwordssecurity operationsSudo use on Linux servers and credential protection need to be checked.developerPrivilege-escalation surfaces on developer machines can also be targets.caseOpenAI model breached government website to pull hospital-bed datasecurity leaderThe ability of AI models to break into systems on their own is a real incident risk.public-sector decision-makerProtection of government websites against AI agents needs to be reconsidered.journalist and researcherThe case provides a basis for examining how AI models behave.influence operationX removes 13,000 accounts tied to suspected foreign influence operations ahead of US midtermspublic-sector decision-makerPlatforms' pre-election moderation is a regulatory issue.journalist and researcherData on the accounts removed by the platform may be a research source.law enforcementMicrosoft and partners dismantle Chinese-language phishing kit ecosystemsecurity leaderThe report shows the scale of targeted phishing ecosystems.security operationsIndicators from the affected phishing kits can be used in email filtering.regulationBundestag passes act implementing the Cyber Resilience ActdeveloperManufacturers of products with digital elements face new conformity obligations.business leaderProduct liability and compliance costs need to be planned for.public-sector decision-makerMember-state implementation may serve as a model for other countries.incidentJapan declares first cyberspace emergency as AI turns ransomware into 'carpet bombing'security leaderMass AI-assisted ransomware campaigns require dedicated preparation.public-sector decision-makerA state-level emergency declaration sets a new precedent.incidentVibetech database exposed 13.7 million credential recordssecurity operationsChecking the leaked credentials against one's own environment is warranted.citizenPasswords for affected accounts should be changed.incidentBitget confirms $387.5 million stolen by hackers using North Korea-style methodssecurity leaderAttacks on crypto exchanges that resemble North Korean methods point to risk across the financial sector.business leaderSafeguarding high-value assets and authentication is critical.researchNew phishing trick targets the AI that reads your inboxsecurity operationsAI assistants' mailbox access should be restricted.developerFor AI features that process email, treating input as untrusted is a design requirement.caseUK minister says attack similar to OpenAI Medicare breach is likelypublic-sector decision-makerThe minister's warning signals that cyber risk is a government priority.caseGoogle DeepMind's Hassabis and Bill Gates warn of AI-driven cyberattackssecurity leaderWarnings from senior leaders reinforce preparedness for AI attacks as a priority.business leaderAI-driven attacks call for cost and risk calculations.law enforcementFBI, DOJ and Meta act against West African scam networkssecurity leaderDismantling the infrastructure of scam campaigns is a useful reference for threat monitoring.citizenAction against scam networks operating through social platforms is part of the broader effort to protect people from fraud.law enforcement16-year-old allegedly led ransomware gang behind hundreds of attackssecurity leaderRansomware groups may include young offenders, which affects the threat model.incidentCybercriminals claim data leaks at Confluent, Deloitte, KFC and MoneyGramsecurity leaderVerifying the attackers' claims and assessing partner risk is also relevant for the customers of the companies concerned.incidentBurger King Japan confirms data breach affecting 11.18 million user recordsbusiness leaderLeaks of consumer databases pose direct reputational and regulatory risk.citizenAffected users' data can be used for misuse and targeted fraud.incidentHack exposes data of millions of California students held by Educational Testing ServicecitizenStudents' personal data was put at risk.reportCrypto hacks near $1.2 billion in third quartersecurity leaderThe scale of attacks on crypto service providers affects financial exposure.incidentTFG London cyberattack disrupts IT systems at Phase Eight, Hobbs and Whistlesbusiness leaderBrands sharing an IT system go down together because of a single incident.incidentEichstätt: cyberattack paralyses city administrationpublic-sector decision-makerOutages of municipal systems also affect local public services.law enforcementAlleged Qilin ransomware leader extradited from Japan to Germanysecurity leaderLaw enforcement action affects the operational risk posed by ransomware groups and the reporting considerations for victims.incidentCenterPoint data breach affects more than 1 million Texans, attorney general report showscitizenData of affected customers may have been leaked.incidentCyberattack at Central Ohio Primary Care may have exposed patient and employee databusiness leaderHealthcare providers are subject to data breach notification obligations.citizenAffected patients and employees should watch for any use of their data.researchAI makes malware evasion easier, but behaviour still gives it awaysecurity operationsBehaviour-based detection remains relevant against AI-generated malware.journalist and researcherThe research examines the balance between offensive and defensive use of AI.reportPhilippines is fifth most targeted country by Chinese state-backed hackerspublic-sector decision-makerThe ranking of targeted countries matters for government defence planning.journalist and researcherThe report is a source of regional targeting statistics.regulationUpdate on the UK Cyber Security and Resilience Billsecurity leaderDevelopments in the obligations should be monitored.business leaderThe proposed legislation may affect providers operating on the UK market.caseMI5 chief warns against overstating Russian interference in Brexit votepublic-sector decision-makerWhen assessing election interference, the emphasis is on evidence-based judgement.journalist and researcherThe MI5 chief's warning is a reference point when judging the scale of Russian influence.law enforcementFBI arrests Canadian cybersecurity expert in ShinyHunters hacking casesecurity leaderDevelopments in the ShinyHunters case bear on the threat picture.journalist and researcherThe arrest also affects the professional community.toolUS Army moves to field AI agents against autonomous cyberattackssecurity leaderThe deployment of agents on the defence side is accelerating.public-sector decision-makerDefensive AI agents raise regulatory and procurement questions.

← Issue 47 · Issue 49 →

Weekly email

One email a week with the main stories from the Agent Age editions. You confirm by email and can unsubscribe in one click.

Almost done: we sent you an email. Open it and press the confirm link. If it does not arrive within a few minutes, check your spam folder.