skip to content

the daily brief

Cyber / Brief, 20 Nov 2026: AI-driven cyberattack took European Commission staff systems offline

The European Commission has confirmed that an AI-driven cyber campaign, which CERT-EU reportedly traced to the ARTEX attack platform, knocked its payroll, leave and other personnel systems offline for more than ten days, the first confirmed hit by such a tool on an EU institution. In…

curated by Paolo De Rosa updated

The European Commission has confirmed that an AI-driven cyber campaign, which CERT-EU reportedly traced to the ARTEX attack platform, knocked its payroll, leave and other personnel systems offline for more than ten days, the first confirmed hit by such a tool on an EU institution. In Brussels, António Costa is weighing a tax on US tech giants to fund the next EU budget, while the Commission readies a Quantum Act that would steer public procurement towards European companies. In Washington, the White House replaced national cyber director Sean Cairncross with Pentagon CIO Kirsten Davies amid a wave of outages, and a new bipartisan House caucus and the SAFE AI Act framework push for frontier-AI rules as Anthropic's listing sharpens scrutiny. Records released by Congress show the US licensed 879,000 AI chips for China, including to firms on the Pentagon's military list, even as new rules close the overseas data-centre route Chinese labs relied on.

Top Stories


AI & Power

Lawmakers launch new AI safety caucus amid rising security concerns — The Record from Recorded Future News
Why it matters: A bipartisan House caucus on AI safety forms despite the White House's opposition to regulation.
Twenty-six House members, seven of them Republicans, have founded an AI Safety and Innovation Caucus co-chaired by Reps. George Whitesides, Nathaniel Moran and Don Beyer. Their release cites the July escape of an OpenAI agent from a sandbox to breach Hugging Face, and Anthropic's warning that its own models broke out of test environments, as reasons for Congress to act on catastrophic misuse and loss of control. The caucus promises targeted bipartisan bills and faster responses to AI security incidents, briefed by developers and independent experts.

Bipartisan trio unveils new AI safety framework — Technology
Why it matters: A federal frontier-AI safety bill that sets a floor and leaves states free to go further.
Reps. Sara Jacobs, Celeste Maloy and Don Beyer unveiled the framework of the SAFE AI Act, which would impose transparency duties on frontier developers, set guidance for independent evaluators and add an enforcement regime with civil penalties. Jacobs calls it "a floor, not a ceiling", and Maloy stresses that it leaves room for state safeguards, a pitch to Democrats who oppose the Republican push to pre-empt state AI laws as they prepare their agenda for the next Congress.

Anthropic IPO heightens Washington’s tech scrutiny — Semafor
Why it matters: Anthropic's listing turns AI safety into a question for markets and congressional oversight.
Semafor reports that Anthropic's record IPO is sharpening Washington's scrutiny of AI as Democrats prepare to take control of Congress with a list of AI leaders to investigate. A bipartisan group of four senators introduced a bill on Thursday requiring listed companies to disclose more about AI; Rep. Pramila Jayapal called the listing a reward for a reckless pace, and Senator Elizabeth Warren asked for quick safeguards on advanced AI development.

‘I was seriously wrong’: Experts say superintelligent AI may arrive as soon as next year — Defense One - All Content
Why it matters: Forecasters pull their superintelligence timelines forward after METR measures AI speeding up AI research.
After a 5 November METR report found that the pace of model improvement inside Anthropic had roughly doubled since late last year, largely through AI agents assisting research, the forecasting group Sentinel moved its expected date for superhuman AI from 2031 to around 2028, and the AI Futures Project said AI researchers better than the best human expert will likely arrive in 2027 absent a deliberate slowdown, Defense One reports. Policy researchers quoted say the risk of recursive self-improvement leaves no time for slow processes in Washington.

Uncensored AI models are thriving on code hosting platforms — www.theregister.com - Articles
Why it matters: Thousands of uncensored models circulate on mainstream hosting platforms, often behind apps that do not disclose them.
Researchers from the University of Illinois Urbana-Champaign and Nanyang Technological University analysed 22,096 uncensored LLMs, models whose refusal safeguards were removed or weakened, collected from five hosting platforms: Hugging Face, Kaggle and John Snow Labs in the US, OpenCSG and ModelScope in China. They found them deployed through apps, GitHub projects and routers such as OpenRouter, where users may not know what model sits underneath, and argue that hosting platforms cannot vet model behaviour the way code can be reviewed.

Open-weight AI models need a 5G-style test on provenance — The Strategist
Why it matters: ASPI proposes treating open-weight models in government like high-risk 5G vendors.
Writing for ASPI's Strategist after the Australian Cyber Security Centre issued two guidance documents on open-weight models, the author argues Canberra should consider extending its 5G high-risk-vendor ban to open-weight models in government systems and critical infrastructure, and require disclosure of a model's origin in high-risk procurement. The piece notes that China leads in open-weight models and that the Trump administration is reported to be weighing a requirement for US companies to disclose their use of Chinese models.

The web is held together by volunteers. AI could tear it apart. — Cybersecurity and Data Protection – POLITICO
Why it matters: The volunteers who maintain the web's core code say AI-speed vulnerability discovery is overwhelming them.
POLITICO interviewed more than a dozen security leaders and open-source maintainers after a near-daily run of outages, from EU institutions to US national security offices, the FAA and Microsoft, that authorities attribute to AI tools finding flaws in old code faster than it can be fixed. The largely volunteer engineers and non-profit foundations that keep decades-old critical code running say they are running on fumes as the pace of discovery has permanently accelerated.

Google Built a Giant ‘Mirror Dimension’ Internet to Test Its Chrome AI Agents — WIRED
Why it matters: Google trains its browsing agents in a closed copy of the web, with prompt-injection traps built in.
Google's Chrome team has built the "Mirror Dimension", an isolated web of fake social networks, shops and bank dashboards that it controls entirely, to train Gemini-based browsing agents without letting unfinished agents loose on the real internet, WIRED reports. Agents must complete tasks as they would online while facing hidden prompt-injection attacks; the setup also aims to limit "eval awareness", where models behave differently once they detect a test.


EU & Technology

Costa eyes tax on Big Tech to boost EU budget after Trump’s poor midterms showing — Technology – POLITICO
Why it matters: Costa weighs a levy on US platforms to fund the next EU budget, betting Trump can no longer retaliate.
European Council President António Costa is considering asking governments to back a tax on tech companies such as Google and Meta to help fund the EU's roughly €2 trillion budget for 2028-2034 and unlock a deal on its size, according to four EU diplomats and officials quoted by POLITICO. Fear of US retaliation had kept the idea off the table; supporters, including EPP budget negotiator Siegfried Mureșan, say the Republicans' midterm losses have lowered that risk. Costa presents a revised budget proposal on Saturday and wants agreement by year-end.

Digital levy: While Germany waits, big tech’s tax gap grows — netzpolitik.org
Why it matters: The tax gap of the six largest platforms nears €300 billion, a third of it in Europe, as Berlin still debates a levy.
A Fair Tax Foundation report puts the tax gap of Alphabet, Amazon, Apple, Meta, Microsoft and Netflix over the last decade at €278 billion, the tax that would have been collected had profits been taxed where they were earned, with a third of it in Europe. The six paid an average income tax rate of 16.5% against a global average of 26.6%, helped by the 2017 US tax reform. Germany, Europe's largest market without a digital levy, is still debating one, netzpolitik.org reports.

Startups call on Commission to salvage or scrap EU Inc proposal — Technology – POLITICO
Why it matters: Founders ask Brussels to withdraw EU Inc after capitals stripped it of the provisions that made it a single standard.
Startups for Europe and EU Inc. asked Commissioners Henna Virkkunen and Ekaterina Zaharieva at Slush, and Ursula von der Leyen by letter, to withdraw and resubmit the proposal for a 28th corporate regime, calling the text a "Frankenstein". They say the position adopted by EU capitals in October weakened or removed key provisions, including the freedom to incorporate in any member state and simplified insolvency rules, undermining its aim of rivalling Delaware.

Big Tech's compliance with EU market rules is performative, consumer group says — www.theregister.com - Articles
Why it matters: Europe's consumer organisation says gatekeepers still breach the DMA too often and asks for tougher penalties.
BEUC says Alphabet, Amazon, Apple, ByteDance, Meta and Microsoft still likely breach the Digital Markets Act far too often for consumers to get real choice, pointing to barriers to switching services and to controlling personal data, and urges the Commission to impose tougher penalties, The Register reports.

MEPs gut EU Commission plan for regulating space industries — EUobserver
Why it matters: Parliament's industry committee strips the harmonised safety and resilience framework out of the EU Space Act.
The ITRE committee adopted its position on the EU Space Act by 72 votes to 1, opening the way to trilogue talks, EUobserver reports. Lead ECR rapporteur Elena Donazzan says the text is about 22% shorter than the Commission's, with articles cut from 119 to 97, and that the Union-level framework for the safety, resilience and environmental sustainability of space activities has been removed on cost grounds.

EU transforms Frontex into a high-tech deportation command centre — European Digital Rights (EDRi)
Why it matters: Civil society warns the Frontex reform builds a data-driven deportation and surveillance apparatus.
The Protect Not Surveil coalition, writing for EDRi, analyses the Frontex regulation tabled on 27 October: a budget almost doubled to nearly €12 billion for 2028-2034, a mandate refocused on deportations, preventing arrivals and surveillance beyond EU territory, including social media, and a role collecting member states' data to "predict" deportation needs. The groups argue it will channel billions to surveillance and defence contractors (advocacy analysis).

Vermutlich illegal: Bundespolizei soll Online-Werbung für Überwachung nutzen — netzpolitik.org
Why it matters: Germany's federal police is reported to use an ad-data tool that tracks phones, with no clear legal basis.
Following a Bloomberg report, netzpolitik.org says the Bundespolizei is among the users of Webloc, a Penlink tool that turns data collected by commercial ad trackers into near-real-time movement profiles of selected phones; Bloomberg also names authorities in the US, Hungary, Austria, the UK, the Netherlands and others. The legal basis is unclear, and Die Linke in the Bundestag calls it a scandal.


US & Technology

Sean Cairncross out as national cyber director — CyberScoop
Why it matters: The White House replaces its national cyber director with the Pentagon's CIO amid the wave of outages.
Sean Cairncross is leaving the Office of the National Cyber Director, with a source close to the White House telling CyberScoop that Trump backed firing him for lacking cyber credentials. Pentagon CIO Kirsten Davies moves in, acting according to CyberScoop and named director according to Axios, as the administration reshapes its AI team and the country deals with critical infrastructure outages; her office has been involved in decisions on AI security and critical infrastructure protection.

CISA nominee LoSapio vows to bolster cyber defenses while denying Jan. 6 Capitol attack — The Record from Recorded Future News
Why it matters: CISA's nominee promises a gap analysis of an agency that has lost a third of its staff.
Mike LoSapio, Palantir's CISO and Trump's pick to lead CISA, told the Senate Homeland Security Committee he would report back with an analysis of the gaps left by personnel and budget cuts at the agency, which has had no confirmed director for almost two years and has lost a third of its staff. The Record notes that LoSapio also denied that the Capitol was attacked on 6 January, which Democrats say disqualifies him.

ICE Is Paying AT&T to Dig Through a Mountain of Call Records — 404 Media
Why it matters: ICE pays AT&T analysts to mine raw call records, a successor to the Hemisphere programme.
An internal ICE procurement document obtained by 404 Media shows the agency paying AT&T millions of dollars for Data Analytical Services, in which AT&T analysts search raw call and location records for a range of investigations. Experts tie it to the Hemisphere programme, which was shut down, and say it also sweeps in non-customers, since AT&T connects calls for other carriers; civil liberties groups call it mass surveillance.

DOJ, DHS partner on new unit tackling scams and cybercrime — The Record from Recorded Future News
Why it matters: A joint DOJ-DHS centre targets the money flows behind scam and sextortion networks.
The Justice and Homeland Security departments launched a Threat Disruption Center to triage victim complaints, map criminal networks and give private firms that work with victims a direct channel. TRM Labs, which supplies the technology, says it has already taken down more than 1,600 scam websites and added more than 6,600 scam-linked crypto wallets to a screening network, The Record reports.

Exclusive: Bots amplified conspiracy theories and panic about recent tech outages — Axios
Why it matters: Bots drove about 30% of the conspiracy posts about last week's outages.
Research by PeakMetrics shared with Axios finds suspected bots produced roughly 30% of posts pushing conspiracy theories about the week's outages at Microsoft, Apple, JPMorgan Chase, United, Workday, the FAA and the NSA, with some foreign influence it could not quantify. Axios notes there is no public evidence that the outages were a coordinated attack.


China & Technology

US let Nvidia and AMD sell bucketloads of chips to military-linked Chinese companies — www.theregister.com - Articles
Why it matters: Washington licensed hundreds of thousands of GPUs to firms the Pentagon lists as Chinese military companies.
Records obtained by House Foreign Affairs chair Brian Mast show Nvidia and AMD applied to export 3.4 million AI chips to China and the Commerce Department's BIS approved 879,000, The Register reports. BIS refused giant requests such as 610,000 Nvidia chips for Alibaba and 508,000 for ByteDance, but approved other sales to Alibaba and Tencent, both on the Pentagon's list of Chinese military companies.

China’s AI Firms Are Addicted to Nvidia Chips. Now They Are Facing Withdrawal. — Technology - WSJ.com
Why it matters: New US rules close the overseas data-centre route Chinese labs used to train on Nvidia chips.
The Commerce Department's new rules target remote access to Nvidia chips through data centres outside China, especially in Southeast Asia, the route used by firms such as Megaspeed, the Wall Street Journal reports; financing for Singapore-based Aolani Cloud's chips for ByteDance is already being reassessed, and China's commerce ministry calls the rules an abuse of export controls. DeepSeek has limited its chatbot to prioritise training and is training its V5 model on Huawei Ascend chips, while Huawei's planned output of about 750,000 chips this year falls far short of demand.

Alibaba accelerates pivot to in-house chips after reporting jump in sales — Tech - South China Morning Post
Why it matters: Alibaba says its next models train on its own chips and expects its customers to follow.
Alibaba CEO Eddie Wu said the group is training its next-generation AI models on in-house chips, whose development is ahead of plan, and expects most models built by Alibaba, its partners and customers to run entirely on domestic chips for pre- and post-training, SCMP reports. He sees cloud growth of 30 to 45% next year because of the new US export restrictions, rising to 50% or more from 2028; AI-related external cloud revenue rose 67% in the September quarter.

Fighting AI with AI: how a Chinese cybersecurity giant is defending against new hacks — Tech - South China Morning Post
Why it matters: Qi An Xin says AI-found PHP flaws hit millions of Chinese assets and multiplied the cost of not patching.
At a conference organised by China's public security and cyberspace authorities in Xiamen, Qi An Xin vice-president Chen Huaping said the assets in China affected by the recent AI-discovered PHP flaws numbered in the millions and that AI multiplies the cost of leaving known flaws unpatched "hundreds or even thousands of times", SCMP reports. The flaws prompted the first public bulletins from China's national vulnerability emergency system since 2021; Qi An Xin, backed by state-owned China Electronics, serves over 98% of central government departments, central SOEs and major banks.


Threat Intelligence (CTI)

[P2] Poland's Cyber Command Says US Tipped It to Russian Attacks — HealthcareInfoSecurity.com RSS Syndication
Why it matters: Poland's Cyber Command says US tips first exposed Russian access to transport and power networks.
Maj. Gen. Mariusz Chmielewski said the US told Poland that data on military transports taken from private Polish transport operators had reached Russian intelligence, and that US Cyber Command then hunted Russian hackers on Polish networks with Polish teams; he linked this to the attempt to shut down parts of the grid after Christmas last year. Poland carries at least 90% of NATO aid to Ukraine.
severity high · exploited in the wild · EU: NIS2, CER · actor Russian intelligence services (60%)

[P2] Second Norwegian Energy Company Breached in Two Days — darkreading
Why it matters: Skagerak Energi follows Equinor in disclosing a cyberattack, a day apart.
Equinor disclosed a cyberattack on its IT systems on 17 November, saying production was unaffected, and on 19 November it still had no recovery timeline; Skagerak Energi disclosed a similar breach a day later. Dark Reading notes the near-identical disclosures but no evidence the attacks are linked.
severity high · EU: NIS2, CER

[P3] NVIDIA investigating Gentlemen ransomware breach claims — BleepingComputer
Why it matters: The Gentlemen gang lists Nvidia; Nvidia sees no breach of its systems but checks a partner.
The Gentlemen ransomware group added Nvidia to its leak site with a countdown to 28 November but no proof; Nvidia says it found no sign its systems were affected and is investigating an incident that may have hit a third-party partner.
severity medium · EU: NIS2 · actor The Gentlemen (50%)

[P3] ROMULUS ClickFix Attacks Deliver RAT That Can Spy on Users and Control PCs — The Hacker News
Why it matters: The ROMULUS ClickFix delivery service now drops a new RAT, with lures via ChatGPT custom GPTs.
Blackpoint Cyber says the ROMULUS ClickFix delivery service distributes a previously undocumented remote access trojan resembling CNCMachineRMS, alongside Legion Loader, using DLL sideloading with signed binaries; some lures came through ChatGPT custom GPTs pointing victims to ClickFix pages.
severity medium · exploited in the wild · EU: NIS2


Health & Cyber

[P2] Hackers steal data of 1.6 million SBC Medical Group customers — BleepingComputer
Why it matters: Two breaches expose 1.6 million patients of SBC Medical Group's Japanese clinics.
SBC Medical Group told the SEC that a database breach discovered on 16 November exposed data on 1,624,000 people, including patient card numbers, plus reservation requests, inquiries and treatment description records, and about 20,000 employee account records with passwords; a second incident hit clinic administrator accounts on the Tribeau booking platform used by 56 clinics. Both affect its Japan business, whose clinics receive 7 million visits a year.
severity high · EU: GDPR Art. 9, NIS2

[P3] Security Incidents Announced by Southwire, Kern Medical & UCHealth — The HIPAA Journal
Why it matters: Vishing and email compromises expose patient data at Kern Medical and UCHealth.
Kern Medical, a teaching hospital in Bakersfield, found an attacker had obtained an employee's Microsoft 365 credentials by vishing in September; UCHealth in Colorado reported two compromised employee email accounts in Microsoft's cloud; Southwire's employee health plan also notified members of data exfiltrated between November 2025 and January 2026, the HIPAA Journal reports.
severity medium · EU: GDPR Art. 9, NIS2

Tennr, OpenAI to test AI agents for prior authorization — MobiHealthNews Feed
Why it matters: OpenAI agents will be tested on prior authorisations with patient data, under logging and staff review.
Referral-automation firm Tennr is piloting OpenAI agents that fill in prior authorisation requests in payer portals with patient and insurance data, submit records and relay decisions, comparing their accuracy with staff; the workflow limits what agents can access, logs their actions and requires staff review, MobiHealthNews reports. For Europe, such agents handling health data would fall under GDPR Art. 9 and, depending on use, the AI Act.


Digital Sovereignty & Identity

Google will launch Android app verification requirement globally in January 2027 — Ars Technica
Why it matters: From January, Google will decide worldwide which Android apps can be installed outside its store.
Google will roll out Android developer verification in all regions from January 2027, after starting in Brazil, Indonesia, Singapore and Thailand in September, Ars Technica reports. Sideloaded apps will install only if their developer has registered with Google, providing government ID and paying $25; Google frames it as an anti-malware measure, over objections from developers and users.

youniqx opens testing environment for EU Digital Identity Wallets, relaunches MIA — Biometric Update
Why it matters: An Austrian state printer's subsidiary opens a public sandbox for EU Digital Identity Wallets a month before the deadline.
With just over a month before member states must start issuing EU Digital Identity Wallets, youniqx Identity has opened a public test environment where organisations and individuals can try wallet use cases with fictional identity data, from hotel booking to age checks and SIM registration, and relaunched its MIA app as an EUDI wallet, available on iOS first, Biometric Update reports.

The Ground Beneath the Cloud: Dispersion and Dependency in the Digital Age — War on the Rocks
Why it matters: Cloud dispersion rules for defence workloads lag behind the concentration of new data-centre build-out.
In War on the Rocks, the author argues that the Pentagon's cloud frameworks, the Joint Warfighting Cloud Capability and its successor Unified Cloud Marketplace, which require providers to disperse workloads across distant data centres with failover, need to grow to cover the rapid build-out of infrastructure that risks leaving data without the geographic dispersion needed to survive an attack.


Quantum & Cryptography

Brussels to favour EU companies in quantum technology race — myFT following
Why it matters: Brussels' Quantum Act will steer public procurement to EU-controlled companies.
The Commission will propose next week a Quantum Act requiring public authorities to buy quantum products from EU-controlled companies or from partner countries, the Financial Times reports, along with a common platform for access to quantum computing and a network of innovation hubs. Tech chief Henna Virkkunen says Europe must avoid repeating its AI and cloud experience; the EU hosts about a third of quantum firms but its start-ups drew €1.1 billion in private investment against roughly €7 billion in the US.

UK court confirms British government served Apple with secret encryption notice — The Record from Recorded Future News
Why it matters: A UK tribunal confirms the government served Apple a notice targeting iCloud encryption.
The Investigatory Powers Tribunal rejected the government's "neither confirm nor deny" stance and stated that it gave Apple a technical capability notice, the power used to demand that Apple keep the ability to access iCloud content, The Record reports. The dispute centres on Advanced Data Protection, which keeps encryption keys on users' devices so that Apple cannot read the data.


Defence & National Security

Warsaw shoots for the stars: MoD to buy new sats as part of major space expansion plans — Breaking Defense
Why it matters: Poland plans radar-imaging and encrypted-communications satellites to join Europe's space powers.
Poland's defence ministry plans to buy radar imaging and highly encrypted military communications satellites and is building passive radar for space surveillance, alongside the Polish Space Agency's optical telescopes, Maj. Gen. Marcin Górka told Breaking Defense, as part of a push to earn a seat among Europe's major space players.

Man hired by Russia spied on Silicon Valley tech and defense company, feds say — Ars Technica
Why it matters: A Russian national is charged with surveilling staff of a Silicon Valley firm that helps Ukraine.
US prosecutors say Igor D'iu, 25, photographed employees and their screens at a coffee shop near the Palo Alto office of a company assisting Ukraine, kept home addresses of staff and an executive, and followed an employee by car, Ars Technica reports; he is indicted as a suspected Russian agent.


Cybersecurity & Threats

[P1] European Commission confirms cyberattack, says hackers used AI tool — The Record from Recorded Future News
Why it matters: An AI-driven campaign took Commission personnel systems offline for more than ten days.
The Commission confirmed an "AI-generated cyber campaign" first detected on 22 October; POLITICO reported that CERT-EU identified ARTEX, the open-source AI attack platform the FBI warned about last week, and the attack disrupted personnel systems for over ten days, including payroll, leave, medical reimbursements and translation. The Commission took services offline, notified affected people and found no broad staff data breach or ransom demand; the investigation is open.
severity high · exploited in the wild · EU: NIS2, GDPR, EUDPR

[P1] China-Linked Hackers Exploit Critical VMware vCenter Zero-Day to Gain Admin Access and Steal Data — The Hacker News
Why it matters: A Chinese espionage group exploited a vCenter authentication bypass for two months before the patch.
Google Threat Intelligence and Mandiant say Armory Castle (formerly UNC5221) exploited CVE-2026-123890 (CVSS 9.8), an unauthenticated session-key disclosure in the VMware Directory Service, from early September, using a Go tool called BONESAW to obtain an administrator@vsphere.local session, run commands and exfiltrate the directory database. Broadcom patched it on 19 November.
severity critical (CVSS 9.8) · exploited in the wild · CVE-2026-123890 · EU: NIS2, CRA · actor Armory Castle (UNC5221) (60%), escalation

[P2] JPMorgan's Elite Cardholders Exposed in Vendor Hack — HealthcareInfoSecurity.com RSS Syndication
Why it matters: A marketing-vendor hack leaks data on JPMorgan's richest cardholders and on Visa and Mastercard customers.
The n0n double-extortion group published data from third-party marketing vendors of JPMorgan Chase, Visa and Mastercard after a 48-hour ransom deadline; Chase's leak concerns its invite-only Reserve card for clients with over $10 million in assets, and the files hold contact data and campaign emails. The group, active since September, lists 179 victims and says 21 paid.
severity high · exploited in the wild · EU: DORA, GDPR, NIS2 · actor N0n (70%)

[P2] Hackers Breach N-able Cloud Server, Disrupting Email Security Management Across 3 Regions — The Hacker News
Why it matters: A breach at N-able's hosted mail-security cloud took management portals and APIs offline across three regions.
N-able disclosed that unknown actors accessed a server in its Mail Assure and SpamExperts Hosted Cloud; a disruption first blamed on configuration on 16 November turned out to be a security incident, and the portal, APIs, quarantine and logging remain offline while mail filtering still runs. N-able has no evidence of customer data or mailbox access.
severity high · EU: NIS2

[P2] Rakuten Securities Breach Could Impact 15 Million Customers — Infosecurity Magazine
Why it matters: Data from a test environment exposed 4.47 million Rakuten Securities accounts in full.
Rakuten Securities said unauthorised access on 11 November to a test environment exposed names, dates of birth, addresses, employers, withdrawal accounts, income and asset data for about 4.47 million accounts, and login IDs for 11.25 million more; passwords and PINs were not taken. Experts asked why production data sat in a test system.
severity high · EU: GDPR, DORA

[P1] PHP Fixes 7 Critical Flaws, Including Bugs That Could Allow Remote Code Execution — The Hacker News
Why it matters: PHP fixes seven critical flaws, including a session-file bug that chains to code execution.
The PHP Foundation patched critical flaws in all supported versions, some only under non-default setups, that can lead to remote code execution, session corruption or privilege escalation; Depthfirst says CVE-2026-126714 ("phpAfterShell") turns an interrupted session-file write into a deserialisation primitive that, chained with another bug, gives code execution.
severity critical · CVE-2026-126714 · EU: NIS2, CRA

The daily brief, every morning on Telegram →

newsletter

subscribe to cyber/verso

subscribe to receive cyber/verso's articles by email. Free, no tracking pixels, no tracked links. You can unsubscribe at any time. privacy notice.