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 23

Apr 20–26. 2026 · 37 stories · 5 of them Hungarian and regionalPage last updated: Nov 21, 2026
Viewpoint

vulnerability · United States

CISA adds eight actively exploited flaws to KEV catalog, including Cisco and JetBrains bugs

On 20 April 2026, CISA added eight new vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, including flaws in Cisco and JetBrains products. Listing in the catalog indicates active exploitation, and according to the reports the agency requires urgent patching.

2026. Apr 20. · 12 sources

Who is affected and how

security leader
The eight flaws must be checked against the organisation's own asset inventory, and those treated as exploited vulnerabilities must be patched out of cycle.
security operations
Cisco and JetBrains instances must be patched or mitigated urgently, and checked for signs of prior exploitation.
jelentés · Egyesült Királyság

UK NCSC chief: most serious attacks on Britain now come from Russia, Iran and China

Speaking at the CYBERUK conference in Glasgow on 21–22 April 2026, the head of the NCSC said the most serious cyberattacks against the United Kingdom now come from Russia, Iran and China. According…

security leaderThe threat assessment justifies prioritizing defense against state actors and preparedness.public-sector decision-makerAccording to the UK authority, preparing for attacks by state actors is a priority, which bears on critical infrastructure protection planning.
incident · global

New self-spreading npm supply-chain attack steals authentication tokens

A new supply-chain attack targeting the npm ecosystem steals developers' credentials and tries to spread further through published packages. Sources such as BleepingComputer reported on the campaign…

security leaderManaging open-source dependencies and token lifetimes is a key risk area.security operationsDependencies and token use in CI/CD environments should be examined for signs of theft and self-propagation.developernpm authentication tokens and package-publishing rights should be checked immediately and rotated if necessary.
vulnerability · global

Microsoft releases out-of-band .NET update to fix severe flaw

On 21 April 2026, Microsoft released an emergency out-of-band update for .NET to fix a severe flaw. Nine sources reported on it.

security operationsThe out-of-band fix requires deployment outside the regular patch cycle.developerRuntimes and builds of applications that use .NET must be upgraded to the released update.
incident · global

Vercel confirms security incident via compromised third-party AI tool

Vercel confirmed on 20 April 2026 that it suffered a security incident caused by the compromise of a third-party AI tool. According to Hungary's National Cybersecurity Institute, the compromise of…

security leaderAccess rights granted to external AI tools, and the integration permissions given to them, should be reviewed.developerTeams using Vercel should check their projects' access rights and secrets.business leaderAI tools in the supply chain pose a direct data-protection exposure.

Hungary

fraudPhishing scams again spread in the name of MVM Next, Hungarian police warnbusiness leaderIt is worth communicating proactively about the authenticity of customer-communication channels and about fraud warnings.citizenText messages and links received in the name of MVM Next should be treated as suspicious and not clicked.reportHungarian election campaign: Szijjártó alleges outside interference, Political Capital analyses Russian attemptspublic-sector decision-makerClaims of influence surrounding the Hungarian election campaign touch on the protection of electoral integrity.journalist and researcherClaims of influence during the campaign period must be separated by source, according to measured activity and political assertion.reportHungarian transfer fraud falls to three-year low, with successful cases down by a thirdpublic-sector decision-makerThe drop in fraud could serve as a basis for assessing the impact of financial protection measures.citizenThe decline points to a lower risk, but changing fraud methods still call for vigilance.regulationPolish Ministry of Digital Affairs explains scope of amendment to national cybersecurity actbusiness leaderCompanies operating on the Polish market need to check whether the amendment applies to them.public-sector decision-makerPolish regulation is also an example of how the cybersecurity legal environment in the region is developing.law enforcementCharges filed against phishing criminals who caused HUF 15 million in damagecitizenThe case shows that phishing scams can cause significant financial harm to Hungarian victims.

More key stories of the week

regulation · United States

NIST moves to risk-based NVD model as CVE submissions rise 263% since 2020

On 20 April 2026, NIST announced it would scale back NVD enrichment and move to a risk-based model, because CVE submissions have risen 263% since 2020. According to Hungary's National Cybersecurity…

security leaderOrganisations will increasingly have to prioritise vulnerabilities and score them themselves.security operationsVulnerability management tools can no longer count on NVD enrichment and scores for every CVE.
vulnerability · global

Critical remote code execution flaw reported in Anthropic's MCP

Reports say a critical vulnerability in Anthropic's Model Context Protocol (MCP) allows remote code execution attacks. Three sources reported the news on 21 April 2026.

security operationsThe exposure and update status of MCP servers and clients must be checked.developerAgent integrations that use MCP should be reviewed as the flaw and its fix come to light.
incident · global

North Korean hackers blamed for roughly $290 million KelpDAO crypto theft

The KelpDAO DeFi platform suffered a crypto theft of about $290–292 million on 20–21 April 2026. CosmicBytez Labs says the Lazarus Group is behind it, other reports blame North Korean hackers, and…

security leaderThe record-sized DeFi theft highlights gaps in platforms' security design.business leaderFor crypto platforms and DeFi service providers, the theft poses liquidity and trust risks.
vulnerability · United States

CISA adds D-Link DIR-823X flaw CVE-2025-29635 to KEV catalog

On 24 April 2026, CISA added the D-Link DIR-823X router flaw, tracked as CVE-2025-29635, to its Known Exploited Vulnerabilities catalog. Lyrie Research and other sources reported the news.

security operationsIf a D-Link DIR-823X is in use in the environment, its replacement or isolation should be considered.citizenUsers of a D-Link DIR-823X at home or in a small office should check whether their device is affected.
incident · European Union

19 million records exposed in breach of French ID agency ANTS

Reports say 19 million records were exposed in an attack on ANTS, the French identity and vehicle-documents agency; the portal was reported as the target of a cyberattack on 21–22 April 2026. The…

public-sector decision-makerProtecting state document systems and breach-notification obligations are key issues.citizenThe leaked personal data can be used against those affected for phishing and identity theft.
vulnerability · United States

CISA flags another SD-WAN vulnerability as actively exploited

On 21 April 2026, CISA flagged another SD-WAN flaw as actively exploited. Six sources reported on the move.

security leaderThe exposure of network edge devices and the logs must be rechecked for signs of exploitation.security operationsSD-WAN device patching must be brought forward, because CISA has classed the flaw as actively exploited.
research · global

AI tools help mediocre North Korean hackers steal millions, report says

According to the report, AI tools are helping mediocre North Korean hackers steal millions. Four sources published the article on 22 April 2026.

security leaderAttackers' use of AI can raise the capabilities of less skilled groups too, which matters for threat modelling.journalist and researcherClaims about offensive use of AI require checking the sources and evidence behind them.
law enforcement · United States

Ransomware negotiator pleads guilty to conspiracy to deploy ransomware and extort US victims

According to a US Department of Justice statement, a Florida man who worked as a ransomware negotiator pleaded guilty to conspiracy to deploy ransomware and extort US victims. The statement was…

security leaderVetting external negotiators involved in incident response for trustworthiness and conflicts of interest is an important consideration.business leaderParticular care is needed when selecting external intermediaries during a ransomware incident.
incident · United States

ADT suffers data breach; ShinyHunters claims it stole 10 million records

ADT suffered a data breach, according to reports up to 24 April 2026. The group ShinyHunters claims it stole 10 million records.

business leaderThe group's public claims of data theft pose reputational and data-protection risks.citizenAffected customers should reckon with the possibility that their data will be misused.
case · United States

Trump's CISA director nominee withdraws after tumultuous nomination process

Donald Trump's nominee for CISA director withdrew on 22 April 2026 after a tumultuous nomination process. Four sources reported the news.

security leaderUncertainty surrounding US federal cybersecurity governance matters also for organizations that rely on CISA guidance.public-sector decision-makerUncertainty over CISA's leadership affects the continuity of US cybersecurity coordination.
incident · global

Euronews: hackers breach Anthropic's Mythos model, which it deemed too dangerous to release

According to a 22 April 2026 Euronews report, hackers breached Anthropic's Mythos AI model, which the company had described as too dangerous to release. Three sources reported the incident; no…

security leaderProtecting access to restricted-availability AI models, and the risk of models leaking, must be managed.journalist and researcherDetails of the report and Anthropic's response should be awaited before the claim is published as fact.
research · global

Mozilla says Anthropic's Mythos found 271 security vulnerabilities in Firefox 150

Mozilla announced on 21 April 2026 that Anthropic's Mythos model found 271 security vulnerabilities in Firefox 150. Three sources carried the announcement.

security leaderUsing AI defensively in vulnerability discovery can speed up the patching cycle.developerAI-based vulnerability discovery can produce a large volume of findings, and handling them burdens the development process.

In brief

vulnerabilityVulnerability reported in Google's Antigravity AI agent managersecurity leaderThe security risk of agent-management tools should be considered when protecting AI development environments.developerDevelopers using Antigravity should follow patch advisories and limit the permissions granted to agents.incidentRansomware attack on healthcare IT provider affects Dutch hospitalssecurity leaderThe compromise of healthcare suppliers directly threatens the operations of the institutions they serve.public-sector decision-makerRisk management for healthcare IT suppliers is also a regulatory question.incidentNew malware tried to sabotage Israel's water system but failed because of buggy codesecurity leaderMalware targeting critical infrastructure poses a risk even when its code is faulty.security operationsProtecting industrial control systems and water-utility OT networks is a priority task.reportInsurers move to cap cyber payouts tied to AI and 'LLMjacking'security leaderInsurers' terms may also require documentation of AI risk management and controls.business leaderCyber insurance may cover AI-related losses less broadly in future.incidentNGate Android malware uses HandyPay NFC app to steal card datasecurity leaderMobile payment fraud remains a focus of banking and payment defences.citizenInstalling unknown NFC apps and entering card details is risky.caseMeta sued over scam ads on Facebook and Instagrampublic-sector decision-makerPlatform liability for scam ads is a regulatory issue.citizenAds on social platforms can carry a risk of fraud.regulationEU imposes sanctions on Russian disinformation organisationspublic-sector decision-makerEU sanctions are part of the toolkit against disinformation operations.journalist and researcherTracking the sanctioned organisations and their activities matters to experts and journalists covering the topic.incidentGerman politicians and journalists may have had Signal messages accessed via phishing; prosecutors investigate espionagepublic-sector decision-makerThe security of politicians' communications can also be a national security issue.journalist and researcherSource protection and securing accounts on encrypted messengers are a priority.reportScramble to protect Britain's undersea cables from sabotagesecurity leaderThe risk of sabotage also matters in planning network resilience.public-sector decision-makerProtecting undersea cables is part of critical infrastructure protection.incidentHackers may have accessed personal data of up to 70,000 people in Canada Life breachbusiness leaderInsurers' customer-data protection and breach-notification obligations are a key risk area.citizenPeople affected need to watch for signs that their data is being misused.incidentCitizens reports limited customer impact after vendor data incident amid ransomware claimssecurity leaderRisk management should include checking third parties' data handling and managing vendor incidents.business leaderRisk in vendors' data handling bears directly on customer trust.incidentThree US healthcare organizations disclose the size of their data breachessecurity leaderHealthcare data breaches signal a frequent risk involving sensitive data.citizenThe impact of leaked health data can be long-lasting.regulationFormer FBI cyber official proposes terrorism designations for hospital ransomware attackssecurity leaderThe regulatory environment for hospital ransomware defence may also change.public-sector decision-makerThe proposal raises the question of how to classify legally ransomware attacks on healthcare and whether to tighten sanctions.incidentSui-based Volo Protocol exploited; team pledges to absorb lossesbusiness leaderOperators of DeFi protocols must reckon with the financial and trust consequences of an exploit.incidentAll residents of Dutch municipality Epe hit by cyberattack, 1,000 ID copies stolenpublic-sector decision-makerProtecting local government data and informing those affected are immediate tasks.citizenStolen ID copies can be used for identity theft and fraud.

← Issue 22 · Issue 24 →

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.