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 19

Mar 23–29. 2026 · 41 stories · 9 of them Hungarian and regionalPage last updated: Nov 21, 2026
Viewpoint

incident · global

TeamPCP supply chain attacks compromise KICS GitHub Action and LiteLLM PyPI packages

According to Wiz's blog, on 24 March 2026 TeamPCP distributed LiteLLM PyPI versions 1.82.7 and 1.82.8 with malicious code, using .pth files for hidden persistence and data exfiltration. Another cluster of reports says the KICS GitHub Action was also compromised, and the Hungarian National Cyber Security Institute issued a warning about the LiteLLM attack.

2026. Mar 23. · 12 sources

Who is affected and how

security leader
Auditing dependencies, including the AI toolchain, and managing CI/CD secrets are priorities.
security operations
Affected environments should be searched for traces of .pth-based persistence and data exfiltration.
developer
Pipelines that use the affected LiteLLM versions or the KICS GitHub Action must be checked, and secrets rotated.
vulnerability · United States

CISA adds F5 BIG-IP APM flaw CVE-2025-53521 to KEV over active exploitation

According to reports of 28 March 2026, CISA added the vulnerability CVE-2025-53521 to its Known Exploited Vulnerabilities catalog. The flaw affects F5 BIG-IP APM, and active exploitation has been…

security leaderWith the KEV listing, patching network edge devices becomes a top priority.security operationsInternet-facing BIG-IP APM instances should be patched immediately and checked for signs of compromise.
vulnerability · United States

CISA: another critical Langflow flaw actively exploited to hijack AI workflows

According to a CISA warning of 26 March 2026, hackers are actively exploiting a critical vulnerability in Langflow, a framework for building AI workflows. The aim is to hijack AI workflows.

security operationsLogs of Langflow instances and the access rights assigned to them should be reviewed.developerAgent and workflow systems built on Langflow should be updated and kept off the public internet.
vulnerability · United States

CISA adds Trivy vulnerability CVE-2026-33634 to Known Exploited Vulnerabilities catalog

According to reports dated 26 March 2026, CISA added the Trivy flaw CVE-2026-33634 to its Known Exploited Vulnerabilities catalog. The coverage describes it as a critical supply chain risk requiring…

security operationsA KEV listing sets a mandatory patching priority for US federal agencies and is a strong signal for others.developerBuild and CI environments that run Trivy must be updated.
vulnerability · global

Active reconnaissance under way for Citrix NetScaler memory overread flaw CVE-2026-3055

According to reports of 28 March 2026, active reconnaissance is under way for CVE-2026-3055 (CVSS 9.3), a memory overread flaw in Citrix NetScaler. The summary does not report exploitation, but the…

security leaderOrganisations should prepare for edge-device risk similar to that seen in earlier NetScaler flaws.security operationsPatching NetScaler devices and monitoring for unusual scanning is urgent.

Hungary

caseHungarian government files espionage charges against prominent investigative journalistjournalist and researcherThe espionage charge poses press and source-protection risks for investigative work.citizenThe case touches on press freedom and surveillance issues.reportPoland faced surge in cyberattacks in 2025, including major attack on energy sectorsecurity leaderEnergy companies in the region should review their OT protection and network segmentation.public-sector decision-makerThe experience of a neighbouring country offers lessons for protecting critical infrastructure.fraudScammers misuse police name; police warn of online fraudcitizenIf an unexpected contact claims to come from an authority, verify it through the official channel.fraudScammers use fake NAV websitesbusiness leaderMisleading messages affecting accounting and finance should be flagged within the company as well.citizenTax-related sites should only be used at their official address, and suspicious links should not be opened.vulnerabilityHungary's NKI issues alert on publicly available spyware affecting iOS 18security operationsiOS 18 devices should be updated and checked through mobile device management.citizenUpdating the operating system on iPhones reduces the risk.influence operationRussian bot network spreads fake news about alleged assassination attempt on Orbánjournalist and researcherIdentifying the false narrative and the traces of the bot network is a fact-checking task.citizenReports of an assassination attempt should not be shared without checking them first.fraudHungary's NKI warns of new phishing campaign using convincing Microsoft-style emailssecurity operationsMessages imitating Microsoft services need to be filtered, and a reporting channel needs to be maintained.citizenLinks in messages that appear to come from Microsoft should be checked before use.caseSzijjártó claims election interference in wiretapping affairpublic-sector decision-makerThe political claim bears on how the wiretapping affair is judged institutionally and in terms of security.journalist and researcherThe evidence behind the claim has to be examined separately to verify it.caseQuantum encryption and future cybersecurity discussed at Budapest meetingsecurity leaderPreparing for post-quantum cryptography is part of long-term planning.

More key stories of the week

law enforcement · United States

Russian initial access broker sentenced to 81 months for aiding ransomware gangs

A US court on 24 March 2026 sentenced a Russian hacker to nearly 6.75 years in prison for providing initial access to ransomware groups. According to the sources, the attacks caused more than $9…

security leaderThe sentence shows that access brokers are a distinct, prosecutable actor in the ransomware chain.business leaderProtecting against initial access, including access rights and credentials, directly reduces business risk.
incident · United States

Iran-linked hackers breach FBI Director Kash Patel's personal email accounts

According to a cluster of reports dated 27 March 2026, Iran-backed hackers gained access to the personal emails of the FBI director. The available headlines and excerpts do not detail the content…

security leaderPersonal accounts of senior officials are also targets for state actors, so they need to be included in the threat model.public-sector decision-makerUsing personal accounts for official communication creates risk.
incident · United States

WorldLeaks claims theft of 160 GB from Los Angeles; Foster City declares emergency

WorldLeaks claims it stole more than 160 GB of data from Los Angeles systems. Foster City, California, declared a state of emergency after a ransomware attack, and city operations were still largely…

public-sector decision-makerProlonged outages of municipal systems affect public services, so a manual fallback operating plan is needed.citizenDisruption to city services and a possible data leak may directly affect residents.
law enforcement · global

Russia arrests alleged administrator of LeakBase cybercrime forum

According to TechCrunch, Russian police on 25 March 2026 detained the alleged administrator of the LeakBase cybercrime forum. Hungarian media, citing the Russian claim, reported on the person they…

security leaderThe forum's removal may shift where data leaks surface, so watchlists should be updated.journalist and researcherThe forum was a known platform for trading leaked data, and changes in its operation should be monitored.
incident · European Union

Dutch Finance Ministry investigates cyber breach of internal systems

The Dutch Ministry of Finance is investigating a cyber intrusion that affected its internal systems, according to reports dated 24 March 2026. Hungarian media also covered the incident.

security operationsThe incident is a timely example of why internal government systems need segmentation and logging.public-sector decision-makerAn attack on finance ministry systems raises questions about the protection of government networks.
report · global

Munich Re: AI makes cyberattacks costlier and more effective

Munich Re said on 25 March 2026 that AI is making cyberattacks costlier and more effective. Another article reports that AI and digital assets are making cyber insurance claims harder to interpret.

security leaderThe risk of AI-assisted attacks must be built into risk management.business leaderCyber insurance terms and risk assessments may change.
incident · United States

Stryker recovers after cyberattack, with manufacturing largely restored

Medical technology company Stryker said the cyberattack has been contained and systems are being restored. According to Reuters, on 26 March operations were improving toward full capacity, after…

security leaderFor healthcare suppliers, the recovery plan and manual fallback operation are critical.business leaderThe attack's impact on business continuity and patient care is part of executive responsibility.citizenIn this case, patients' surgeries were delayed.
attribution · United States

FBI says Iranian hackers use Telegram to exfiltrate data in malware attacks

The FBI says Iranian state-linked hackers are using Telegram to exfiltrate data stolen from infected machines. The warning was published on 23 March 2026.

security operationsMonitoring and filtering outbound traffic to Telegram can be a detection opportunity.journalist and researcherThe exfiltration channel described by the authority offers a new lead for analysis and detection.
incident · global

Railway PaaS platform abused in Microsoft 365 device code phishing campaign

The Railway platform-as-a-service was used in a mass device code phishing campaign targeting Microsoft 365, according to reports dated 24 March 2026. The attackers relied on deceiving users in the…

security operationsRestricting device code authentication and monitoring unusual sign-ins reduces the risk.citizenUnexpected device code approval requests should not be accepted.
vulnerability · global

Oracle issues emergency patch for critical Identity Manager vulnerability

On 23 March 2026, Oracle released an urgent patch for a critical vulnerability in Identity Manager. The summary does not give details of the flaw.

security leaderCompromise of identity management systems can give broad access, so they are a priority risk.security operationsOrganisations running Identity Manager must patch out of cycle.
incident · global

AI-assisted phishing campaign compromises hundreds of organizations

An AI-assisted phishing campaign compromised hundreds of organizations, according to a report dated 23 March 2026. The summary does not name the attacker or detail the method.

security operationsAI-generated scam messages are harder to spot, so stronger technical filtering is needed.business leaderStaff training should be refreshed with realistic examples of AI-driven phishing.
report · United States

ODNI report: China, Russia, Iran and North Korea pose growing cyber risks to US critical infrastructure

A report by the Office of the Director of National Intelligence (ODNI) says US critical infrastructure faces growing cyber risks from China, Russia, Iran and North Korea. The report was published on…

security leaderIn sectors targeted by state actors, the threat model needs to be updated.public-sector decision-makerThe government assessment provides guidance for setting critical infrastructure protection priorities.

In brief

law enforcementEuropol-backed operation takes down more than 373,000 dark web sitespublic-sector decision-makerInternational cooperation serves as a model for law enforcement.journalist and researcherThe operation's impact on the dark web ecosystem should be tracked.vulnerabilityClaude extension flaw enabled zero-click XSS and prompt injection via any websitesecurity operationsThe enabling of AI assistants integrated into the browser should be governed by policy.developerInput handling in AI extensions and tighter browser permissions are critical.researchResearchers warn of vulnerabilities in AI-generated codesecurity leaderUsing code-generation tools requires a security review process.developerAI-written code must be audited and tested in the same way as human-written code.researchEmergency department AI gives out methamphetamine recipe in jailbreak testingsecurity leaderJailbreak testing is mandatory before deploying AI systems in healthcare.public-sector decision-makerClinical AI tools need safety requirements and oversight.regulationLiberal government targets deepfakes and foreign interference in major election law reformspublic-sector decision-makerElection rules on deepfakes could serve as a legislative model for other countries.journalist and researcherRules on labelling deepfakes and handling interference also affect fact-checking.report32% of the most exploited vulnerabilities are more than a decade oldsecurity leaderOld flaws deserve separate attention in the patch-management priority order.security operationsInventorying and patching old, unpatched systems remains critical.incidentCalifornia semiconductor testing firm Trio-Tech International hit by ransomware attacksecurity leaderSegmenting the production networks of manufacturing and testing companies is critical.business leaderAs a listed company, it must inform the regulator of a ransomware incident within the deadline.incidentAstraZeneca reportedly hit by data breachsecurity leaderPharmaceutical firms should review their data connections with partners and suppliers.business leaderThe regulatory and business consequences of data breaches are particularly significant in the pharmaceutical industry.reportRecruits reach Russian sabotage training camps for $500 and a trip abroadpublic-sector decision-makerSabotage recruitment also bears on the physical protection of critical infrastructure.journalist and researcherRecruitment methods can serve as sources for research on hybrid threats.incidentMazda discloses security breach exposing employee and partner datasecurity leaderThe scope of data shared with partners needs to be reviewed.business leaderIn the automotive supply chain, exposure of partner data is a shared risk.incidentAnthropic's upcoming Claude Mythos model leaked in security incidentsecurity leaderThe security practices of AI providers are part of supplier risk assessment.journalist and researcherThe exposure of AI developers' internal materials is a risk in the AI supply chain.vulnerabilityChrome 146 update patches high-severity vulnerabilitiessecurity operationsUpdating the browser fleet should be accelerated.incidentNavia data breach affects 2.7 million peoplebusiness leaderHandling sensitive data on a large number of people brings a significant notification and legal burden.citizenThose affected should watch for signs of misuse of their personal data.incidentCrunchyroll investigates hacker's claim of stealing data of 6.8 million userssecurity leaderVerifying attackers' claims and communicating during incident response are shared tasks.citizenUsers should change their passwords and expect phishing attempts.regulationVietnam's National Assembly passes new cybersecurity lawbusiness leaderCompanies operating in the Vietnamese market must examine the new law's compliance requirements.

← Issue 18 · Issue 20 →

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.