the daily brief
Cyber / Brief, 15 Nov 2026: Justice Department probes widen around Nvidia after Huang's exit
The fallout from Jensen Huang's exit is moving from Nvidia's boardroom into Washington's machinery: the Justice Department has opened inquiries touching Nvidia, Megaspeed and Supermicro, Huang has agreed to testify to the House, and a Wall Street Journal investigation shows how Malaysian…
curated by Paolo De Rosa updated
The fallout from Jensen Huang's exit is moving from Nvidia's boardroom into Washington's machinery: the Justice Department has opened inquiries touching Nvidia, Megaspeed and Supermicro, Huang has agreed to testify to the House, and a Wall Street Journal investigation shows how Malaysian and Singaporean data centres kept China's AI labs supplied, a route Commerce now wants to close with licences for Southeast Asia. In Europe the sovereignty argument found its own evidence, as a day-long Microsoft outage hit the Dublin leaders' meeting that ended with Ursula von der Leyen promising a funding plan for tech sovereignty, while the Bundestag gave Germany's intelligence services the power to hack and sabotage abroad. The threat picture sharpened too: the FBI warned that AI agent frameworks such as Artex now run intrusions at machine speed, a tampered software update grounded drones flown by the Bundeswehr, and Estonia will close its Russian border after blaming the GRU for a sabotage attempt on a drone maker.
Top Stories
- US DOJ probes Nvidia as fallout from Huang’s resignation ripples across Washington — Semafor · AI & Power
- How Jensen Huang Built the Backdoor Route That Keeps Beijing in the AI Race — Technology - WSJ.com · AI & Power
- German parliament passes law giving spy agencies hacking, sabotage powers — The Record from Recorded Future News · EU & Technology
- FBI flash alert says Artex tool automating cyberattacks at ‘machine speed’ — The Record from Recorded Future News · Threat Intelligence (CTI)
- Von der Leyen will present ‘ambitious proposal’ to fund tech sovereignty — Tech | Rapporteur · EU & Technology
AI & Power
US DOJ probes Nvidia as fallout from Huang’s resignation ripples across Washington — Semafor
Why it matters: The Huang affair moves from the boardroom to the Justice Department and Congress.
Two administration officials told Semafor that several Justice Department inquiries could reach Nvidia: one, on antitrust grounds, targets the chipmaker itself, while two others look at Megaspeed and Supermicro and at the role of Alice Huang, the alleged Chinese intelligence asset with whom Jensen Huang admitted an affair. Nothing suggests criminal charges against Nvidia or Huang for now; Supermicro says Megaspeed was never its customer. Separately, POLITICO reports that Huang has agreed to testify before House lawmakers, after the Republican chairs of the Foreign Affairs, China and Intelligence committees asked him to appear and John Moolenaar said a subpoena was not ruled out.
How Jensen Huang Built the Backdoor Route That Keeps Beijing in the AI Race — Technology - WSJ.com
Why it matters: The data-centre build-out in Southeast Asia that kept China's AI labs supplied is now the target.
A Wall Street Journal investigation from Malaysia describes how Nvidia's network of Asian cloud partners rented top chips to Chinese firms: records seen by the paper show Megaspeed arranged to send nearly $12 billion of Nvidia chips to companies in China between March 2024 and June 2025 (delivery and origin unclear), ByteDance anchors Aolani's and DayOne's capacity, and Alibaba plans to use a BlackRock-backed site near Kuala Lumpur. The Commerce Department is drafting rules that would require licences for Blackwell-class exports to Malaysia, Indonesia and Singapore and bar Chinese remote access; Malaysia says it is investigating Megaspeed's local affiliate. Cloud providers are already looking to Japan, Brazil, Central Asia and Europe for new capacity to serve Chinese customers.
Chinese AI labs are training models on Nvidia's Blackwell despite export controls, says ex-OpenAI engineer — www.theregister.com - Articles
Why it matters: A forensic method claims to read the training GPU from a model's weights.
Former OpenAI engineer Will DePue published evidence that DeepSeek, Moonshot AI, MiniMax and Ant Group trained models on Nvidia Blackwell GPUs despite export controls, based on what he describes as a hardware fingerprint left in a model's initial weights that survives training. The Register notes that four bills to limit China's access, including the AI Overwatch, Chip Security and MATCH acts, await Congress, and that Lava recently found 32 B300 GPUs behind monitoring endpoints geolocated to China. The method has not been independently validated.
Can Chinese AI companies curb cyber risks from open-weight models without losing their edge? — Tech - South China Morning Post
Why it matters: China's open-weight labs strip hacking skills from their models instead of closing them.
Tencent says it removed offensive cyber capabilities from its new Hy4 model while keeping defensive ones, after Moonshot did the same with Kimi K3.1 Pro in October, SCMP reports: since open weights cannot be monitored after release, the firms filter attack-related data out of training. Anthropic reported in September that Z.ai's GLM-5.3 could develop sophisticated exploits on its own. Brookings' Kyle Chan says the firms want models that are powerful, open and safe at once and will face trade-offs.
White House AI Task Force Urges Leniency on Cyberattack Penalties — DataBreaches.Net
Why it matters: Washington proposes a year without penalties for companies whose defences AI overwhelms.
After its second meeting on Thursday, the White House AI task force led by Director of National Intelligence Jay Clayton urged federal regulators not to take enforcement action against companies hit by AI-driven cyberattacks for at least 12 months, with possible extension, and to encourage voluntary reporting of vulnerabilities and incidents. Semafor reports that members heard about the common industry standards being drafted by signatories of the White House AI accord and about NIST's pre-release model testing, and that the task force wants Congress to reauthorise cyber information-sharing laws; the Nvidia affair was not discussed.
Companies Linked to Alice Huang Invested in DeepSeek — The Information
Why it matters: The Megaspeed affair reaches China's best-known AI lab through its investors.
The Information reports that companies tied to Alice Huang, Century Litong and its Shanghai-listed parent Jiangsu Lettall Electronics, recently invested in DeepSeek. No readable source carried the story. Excerpt-based.
Open-source AI without Jensen — Interconnects AI
Why it matters: The US open-model ecosystem loses its most powerful patron.
Nathan Lambert argues that Huang's exit is a major setback for American open-weight models: Nvidia still has a commercial interest in them, to avoid dependence on a few customers who build their own chips, but its open-model push was a founder's bet, its funding is likely to dry up, and Huang used his political weight to protect the ecosystem. With compute scarce and open models served at high margins, he writes, the closed labs with integrated chips and data centres are favoured.
Independent AI auditors have a recruitment problem — Transformer
Why it matters: Third-party oversight of frontier AI may fail for lack of people able to do it.
Transformer reports that proposals for independent verification organisations, auditors with advance access to models such as the regime proposed in Massachusetts, face a shortage of staff with the technical and governance skills required, and that training them could take longer than advocates assume. The piece compares the model to securities audits, FDA approval and the Pentagon's third-party certification of contractors.
Introducing ChatGPT Wallet — OpenAI News
Why it matters: Agents get a way to pay, with the card networks inside the loop.
OpenAI launched ChatGPT Wallet, built with American Express, Mastercard and Visa, which lets agents complete purchases with a network payment token instead of the user's card details, after the user approves each purchase with a biometric check; it starts for ChatGPT Pro users in the US. OpenAI expects users later to pre-authorise purchases within spending limits and categories, and published guidance for merchants whose systems may block agents.
EU & Technology
German parliament passes law giving spy agencies hacking, sabotage powers — The Record from Recorded Future News
Why it matters: Germany's largest post-war intelligence reform gives its services offensive cyber powers.
The Bundestag approved by 308 votes to 190, with 71 abstentions, a law that lets the BND hack foreign systems, sabotage adversaries' supply chains and take down servers run by hostile state hackers, and gives the domestic BfV narrower covert powers such as blocking or rerouting traffic, though no use of force against people. Both agencies gain a legal basis to analyse large datasets with AI and match images biometrically against online material. Interior Minister Alexander Dobrindt called it the reform of the century; 28 civil-society organisations had asked MPs to reject it, and netzpolitik.org argues the powers will not make Germany safer.
Von der Leyen will present ‘ambitious proposal’ to fund tech sovereignty — Tech | Rapporteur
Why it matters: The Commission promises money behind the sovereignty agenda at the next summit.
Closing the leaders' meeting in Dublin, Ursula von der Leyen said she would present an ambitious proposal to fund Europe's tech sovereignty when leaders meet again on 26 and 27 November, Rapporteur reports; the Commission already outlined a strategic tech equity capacity this year to keep scale-ups in Europe, and one expert suggested starting at €10 billion, at a time when governments disagree over the next long-term budget. Excerpt-based.
Microsoft outage renews European calls to detox from US tech — Technology – POLITICO
Why it matters: A US provider's 24-hour failure lands in the middle of the EU's sovereignty debate.
The Microsoft 365 outage, which lasted nearly a day in 74 countries, coincided with the leaders' AI meeting in Dublin, where Mistral's Arthur Mensch told them only 20% of European digital services run on European infrastructure and Emmanuel Macron wrote that Europe's independence cannot be outsourced. POLITICO reports separately that the Commission, the Council and the Parliament were largely unaffected; the Commission spoke of minor slowness in SharePoint, and Microsoft said it saw no evidence of malicious activity.
Italy hits Meta with €805m fine over privacy law violation — Technology – POLITICO
Why it matters: A national regulator bypasses the one-stop shop to levy its largest fine.
Italy's data protection authority fined Meta €805 million, by far its largest penalty, because Facebook's camera-roll suggestions on Android read users' photo and video galleries under a single app-wide permission, without specific consent, in breach of the Italian law implementing the ePrivacy directive. The Garante rejected Meta's claim that Ireland was competent, since ePrivacy has no one-stop-shop mechanism, and gave the company 60 days to seek separate consent and disable the feature by default. Meta can appeal.
EU deal on Ukraine talks overshadows AI-focused summit in Dublin — Technology – POLITICO
Why it matters: The first leaders' summit on AI ends with a decision on Ukraine instead.
EU leaders meeting in Dublin spent most of their informal summit on who should represent Europe in Ukraine peace talks, agreeing on the E5 (Germany, France, the UK, Italy and Poland) plus the Commission and the Council, POLITICO reports. The Irish presidency had billed the meeting as a high-level discussion on developing AI according to European values; Foreign Policy notes the bloc is expected to miss the Chips Act's goal of 20% of world semiconductor production by 2030.
AI is making the EU institutions hallucinate — Tech | Rapporteur
Why it matters: The institutions regulating AI are swamped by AI-made paperwork.
Parliament secretary-general Alessandro Chiocchetti told MEPs that amendments, questions to the Commission, petitions and access-to-documents requests have surged because of AI, and Roberta Metsola warned that AI-drafted amendments cannot replace debate, Rapporteur reports. Translators have found references to non-existent laws in MEPs' texts, and the legal service has spent days checking AI-generated claims; Irish MEP Barry Andrews says he is the first to adopt a personal AI policy.
Bundeshaushalt: Bund verschleiert KI-Ausgaben — netzpolitik.org
Why it matters: Berlin wants to be an 'AI nation' by 2030 without saying what it spends.
netzpolitik.org finds that the federal AI task force's report sets targets for making Germany an AI nation by 2030 but states that they justify no funding decisions, while the 2027 budget does not show how much goes to AI: the think tank Agora Digitale Transformation and ZEW count €20.1 billion of digital spending in the core budget plus €4.1 billion from the special fund, and the digital ministry can claim €5.3 billion in total.
Tinder owner hit with €11M privacy fine — Cybersecurity and Data Protection – POLITICO
Why it matters: An automated 'download my data' tool fails the GDPR access test.
Ireland's Data Protection Commission fined Match Group €11 million and ordered it to comply within three months, finding that Tinder's Download My Data tool omitted categories of personal data or gave only summaries, in breach of the right of access; the inquiry opened in January 2020. Match Group will appeal.
US & Technology
Bessent puts Megaspeed heat on Lutnick: "Commerce has the pen" — Axios
Why it matters: The Treasury points at Commerce for leaving Megaspeed off the Entity List.
Treasury Secretary Scott Bessent told CNBC that Treasury, the Pentagon, USTR and State had favoured putting Megaspeed on the Entity List for months but that the Commerce Department 'has the pen', without denying reports of tension with Commerce Secretary Howard Lutnick, Axios reports. The Wall Street Journal says Lutnick's light-touch line on chip exports is losing support inside the administration, while the White House says Trump has full confidence in him. Excerpt-based.
Jensen Out, MATCH Act In? — ChinaTalk
Why it matters: With Nvidia's lobbying weakened, the next fight is over lithography tools, a European industry.
On ChinaTalk, SemiAnalysis' Dylan Patel argues that Huang's exit removes the main opponent of cutting China off and could give cover to poorly designed laws: he backs the MATCH Act, which would bar advanced DUV lithography tools from China as EUV already is, and calls the Remote Access Security Act, which targets cloud access through Malaysia, a mistake that would push China towards smuggling and domestic production. inference: MATCH would bear directly on ASML and the Dutch export regime.
Beyond the spy scandal: What the Nvidia-Megaspeed case reveals about export controls — Atlantic Council
Why it matters: The case tests how far Washington can rely on companies to police their own exports.
The Atlantic Council notes that Megaspeed became Nvidia's largest Southeast Asian buyer, importing at least $4.6 billion of Nvidia technology in 2023-2025 and renting chips to Chinese firms such as Alibaba, which current rules allow, while suspected of physically diverting restricted chips, which it denies; no charges have been filed. It argues the government must help the private sector protect sensitive technology against foreign intelligence.
NYC to bar surveillance pricing as momentum for bans grows nationwide — The Record from Recorded Future News
Why it matters: Personalised prices based on personal data are banned in a second US city.
The New York City Council voted 44 to 5 to bar businesses from setting different prices for the same product from customers' personal data, with fines of up to $1,000 per violation and exemptions for transparent discounts and loyalty schemes, after Seattle passed a similar ban last month. A state law passed in June still awaits Governor Kathy Hochul's signature.
China & Technology
China could tighten rare earths exports if US escalates chip curbs over Nvidia: analysts — Tech - South China Morning Post
Why it matters: Beijing's analysts signal the lever China would use if the Nvidia affair hardens US controls.
Chinese analysts told SCMP that Beijing could respond with tighter rare-earth export controls if Washington uses the Nvidia affair to restrict chip sales further, though they expect trade talks to stay on track before next week's meeting between the two leaders. Victor Gao of the Centre for China and Globalisation said using the scandal to pressure Beijing would be a miscalculation. Beijing denies that Alice Huang was a spy.
Nvidia Sought Taipei’s Help Over Employee’s Chip Smuggling Case — Bloomberg Technology
Why it matters: Nvidia lobbied Taiwan's officials over a manager charged in a smuggling ring.
Bloomberg reports that Nvidia raised concerns with Taiwanese political officials about the July arrest of a manager surnamed Chang, indicted in August for forgery and breach of trust, whom prosecutors call the core figure of a network that allegedly smuggled 74 servers with Nvidia chips to China and tried to divert 56 more seized by the coast guard in May. Only the opening of the article was readable. Excerpt-based.
Alibaba’s Chip Unit Wants to Go Public. Who Pays It? — TechNode
Why it matters: China's largest in-house AI chip business seeks public money.
Alibaba confirmed that T-Head, maker of the Zhenwu AI processors and Yitian server CPUs and licensor of Xuantie RISC-V designs, filed confidentially in September for a Hong Kong listing that could raise at least $2 billion; valuations reported range from $25 billion to $85 billion. TechNode notes that more than 60% of its computing capacity serves customers outside the group, mostly through Alibaba Cloud, which makes the chip unit's own revenue hard to separate.
CXMT Moves DDR5 to 32Gb Dies and Lists MRDIMM Server Memory at Up to 12,800Mbps — Pandaily - China Tech News, AI & Electric Vehicle Insights
Why it matters: China's memory champion closes part of the gap in server DRAM.
ChangXin Memory Technologies has moved its DDR5 dies from 24Gb to 32Gb and lists RDIMM and MRDIMM server modules at up to 12,800 Mbps, Pandaily reports. Excerpt-based.
ADA Space Rolls Out Five Satellites for Batch 03 of Its Star-Compute Space Computing Program — Pandaily - China Tech News, AI & Electric Vehicle Insights
Why it matters: China keeps building computing capacity in orbit.
ADA Space rolled out five computing satellites for the third batch of its Star-Compute space computing constellation in Taizhou, Zhejiang, on 13 November, Pandaily reports. Excerpt-based.
Health & Cyber
Crackdown on UK medical record snooping after staff reported sneaking 300+ times this year — www.theregister.com - Articles
Why it matters: The UK regulator asks health executives to vouch personally for access controls.
The ICO wrote to more than 200 senior health leaders after receiving over 300 reports of staff snooping in medical records between January and mid-October 2026, from looking up acquaintances to sharing information, and asked them to confirm safeguards are in place; the figure counts reports by organisations, not confirmed breaches.
[P2] Metabase Zero-Day Exploited in Dorsata Data Breach Affecting 58,000 Patients — The HIPAA Journal
Why it matters: A women's-health records vendor is breached through a business-intelligence tool.
Dorsata, a Virginia maker of electronic medical records for women's-health providers, reported that an attacker exploited a then-unknown Metabase vulnerability between 10 and 12 August 2026 to reach a database of its provider partners' patients, exposing data on 58,249 people. The HIPAA Journal gives no CVE; inference: a Metabase flaw (CVE-2026-72898) entered CISA's KEV catalogue on 11 August.
severity high · exploited in the wild · EPSS 0.985 · CVE-2026-72898 · EU: GDPR Art. 9, NIS2, EHDS
Doctors want liability shield from AI mistakes — Axios
Why it matters: US physicians ask to be shielded when clinical AI harms patients.
The American Medical Association called for federal and state laws protecting physicians from liability when AI products used as intended injure patients, Axios reports, as courts increasingly hold AI developers responsible and the FDA weighs a new framework for AI-enabled devices. Excerpt-based.
Challenges for human oversight in detecting system malfunctions in AI-assisted medical diagnostics — npj Digital Medicine
Why it matters: Doctors as a safety net for clinical AI fail when the AI degrades.
In a randomised study of 48 physicians using an increasingly unreliable AI for cardiac auscultation, published in npj Digital Medicine, diagnostic performance fell with the AI's reliability even when doctors classified first and when they outperformed the AI; their confidence fell faster than their trust in the system. The authors call for defined oversight goals, relevant to the AI Act's human-oversight requirement for high-risk medical AI.
[P2] 🏴☠️ Panzer has just published a new victim : Ministry of Health Malaysia — Ransomware.live RSS Feed
Why it matters: A ransomware group claims a national health ministry.
The Panzer group listed Malaysia's Ministry of Health on its leak site, according to ransomware.live; the ministry has not confirmed an attack and no impact on care has been reported. The same tracker shows n0n claiming Danish drugmaker LEO Pharma.
severity high · EU: NIS2, GDPR Art. 9 · actor Panzer (40%)
Palantir's NHS cancer app gets plenty of praise, precious few users — www.theregister.com - Articles
Why it matters: The flagship use case of the NHS data platform has few users.
Official figures put uptake of Cancer 360, the cancer-pathway app on Palantir's £330 million Federated Data Platform, at five or six NHS trusts, and Freedom of Information responses gathered by No Palantir in the NHS show gaps between reported benefits and recorded use.
Threat Intelligence (CTI)
[P1] FBI flash alert says Artex tool automating cyberattacks at ‘machine speed’ — The Record from Recorded Future News
Why it matters: The FBI confirms AI agent frameworks now run intrusions for criminals and states.
An FBI flash alert says the open-source AI penetration-testing framework Artex is used by criminals and nation-states against government agencies and dozens of critical sectors, chaining reconnaissance to exfiltration at machine speed, in some cases reaching privileged access within three hours; it survives the developer's removal of the repository through forks and a commercial version, and Censys counts over 1,600 exposed consoles. Cisco Talos tracks a wider family of such frameworks, including PentAGI, with over 700 reachable instances, and Hephaestus, controlled through Telegram.
severity high · exploited in the wild · EU: NIS2, AI Act, CRA
[P1] German military drone operations disrupted by cyberattack — BleepingComputer
Why it matters: A tampered update grounds drones used by the Bundeswehr.
Quantum Systems, whose drones serve the German armed forces and which Germany funds to build interceptors for Ukraine, says hostile actors attacked it in late October; Der Spiegel and the FT report a malicious update to its QBase mission-planning software was delivered from the update server, and the Bundeswehr grounded the drones as a precaution. Military counterintelligence (MAD), the BfV and the BSI are involved; flights have resumed.
severity critical · exploited in the wild · EU: NIS2, CRA, EDIRPA
[P2] Ransomware at the hypervisor layer: how attackers use ESXi and vCenter features for malicious purposes — Securelist
Why it matters: A ransomware crew destroys a whole virtual estate through legitimate VMware tools.
Kaspersky's emergency response team describes a BlackNevas intrusion that destroyed four ESXi hosts, every VM on them, both vCenter appliances and all connected backups over ten days, using only legitimate VMware management interfaces and no hypervisor exploit; it notes at least ten families now build on leaked Babuk ESXi code and that victims often run unsupported ESXi versions.
severity high · exploited in the wild · EU: NIS2, DORA · actor BlackNevas (70%)
[P3] Police arrest 354 suspects in dark web cybercrime crackdown — BleepingComputer
Why it matters: A 14-country operation hits vendors and buyers on five dark-web markets.
Operation TorNado, run with Europol from May 2025 to October 2026 by agencies of 14 countries, led to 354 arrests of vendors and buyers on Archetyp, Nemesis, Tor2Door, Bohemia and Kingdom, and seizures of over €60 million and 57 firearms in Europe and over €150 million and 152 firearms in the US.
severity medium · exploited in the wild · EU: NIS2
[P3] ArcticRAT Cybercrime Gang Profiles Victim Finances With AI — darkreading
Why it matters: Criminals use LLMs to sift stolen phone data for money.
Zimperium says the Android stealer ArcticRAT has become a full remote-access Trojan with real-time control hidden behind a black, un-dismissable overlay, and that its operators run stolen SMS and other data through LLM services to find financial information for fraud.
severity medium · exploited in the wild · EU: PSD2, NIS2
Defence & National Security
Estonia to close border with Russia after ‘sabotage attack’ on defense firm — Policy – POLITICO
Why it matters: A NATO state shuts its border over a sabotage attempt on its drone industry.
Estonia will close all crossings with Russia from 1 December after its Internal Security Service said Russia's GRU commissioned an attempted arson and explosion at drone maker Threod Systems near Tallinn early on Saturday; security staff stopped the intruders, and eight suspects, all residents of Latvia, were detained.
Industry guidelines for offensive cyber tools unveiled through international Pall Mall Process — CyberScoop
Why it matters: Spyware vendors get voluntary rules from the UK- and France-led Pall Mall Process.
The Pall Mall Process finalised voluntary guidelines for companies selling commercial cyber intrusion capabilities such as spyware, built on compliance, due diligence, risk management and monitoring, and the United States signed on; Access Now said the compromise among states, industry and civil society falls short of what rights groups sought.
Pentagon rebrands principal AI shop as ‘Super Intelligence Office’ — Breaking Defense
Why it matters: The Pentagon's AI office takes the administration's new vocabulary.
The Defense Department renamed its Chief Digital and Artificial Intelligence Office the Super Intelligence Office, in line with Trump's order to replace 'artificial intelligence' with 'super intelligence' across government; its head Cameron Stanley said the mission is to harness 'sovereign Super Intelligence' for military superiority. The department says the mission is unchanged.
US Army names officer held on ‘allegations of espionage’ — Defense News
Why it matters: A US officer at a NATO corps base in England is held for allegedly helping Russian intelligence.
The Army identified Lt. Col. Brandon Carroll Tyner, assigned to Imjin Barracks, home of NATO's Allied Rapid Reaction Corps, as the soldier arrested in the UK under the National Security Act for reportedly assisting Russian intelligence; FBI Director Kash Patel said the bureau has made 167 counterintelligence and espionage arrests in 21 months.
Digital Sovereignty & Identity
Singapore explores internet domain system to identify and track AI agents — Biometric Update
Why it matters: A state tries to give every AI agent an accountable human behind it.
Singapore's IMDA is working with GoDaddy, Infoblox, Cloudflare and Google on using the domain name system to identify and verify AI agents across platforms, after a Chain of Trust demonstration in October, and is weighing a national registry; an MP proposed one modelled on the Singpass digital identity. Minister of State Jasmin Lau said agents can act for people but cannot answer for them.
Palantir adds three UK police forces to its beat with £10.1M deal — www.theregister.com - Articles
Why it matters: Palantir gains UK police data after London blocked a Met deal.
Bedfordshire, Cambridgeshire and Hertfordshire police awarded Palantir a five-year £10.1 million contract for a platform that ingests and links data across policing systems, after London's mayor blocked a separate deal with the Metropolitan Police. Former Met commissioner Cressida Dick has been a paid Palantir adviser since September 2025; she says she never lobbied for the company.
Google’s AI glasses raise questions about enforcing biometric privacy protections — Biometric Update
Why it matters: A ban on facial recognition is only as good as its enforcement on the device.
Google's AI glasses, made with Warby Parker, Gentle Monster and Samsung and on sale from 10 December, redact bystanders' faces on the paired phone before images reach Gemini and forbid biometric identification by apps, but Biometric Update finds the developer documentation does not say whether the system blocks facial recognition technically or polices apps after they gain camera access; Gemini can analyse surroundings without lighting the recording indicator.
Isabel Schnabel: Made in Europe — ECB - European Central Bank
Why it matters: The ECB's case for a single legal regime for European firms.
ECB board member Isabel Schnabel argued that Europe lacks scale rather than talent or research, and that a '28th regime', a single legal framework giving firms access to the whole single market, would be a game changer and help build a 'Made in Europe' brand.
Quantum & Cryptography
Getting Past the Post-Quantum Cryptography Discovery Phase — darkreading
Why it matters: Most organisations cannot yet say where their cryptography is.
Dark Reading cites a July DigiCert study in which 87% of organisations say they are planning, testing or implementing post-quantum cryptography but only 44% have transition plans or inventories, and a Ponemon study in which 43% have full visibility of their cryptographic estate; practitioners describe companies as stuck in the discovery phase.
Addressing Quantum Readiness in Healthcare Security — HealthcareInfoSecurity.com RSS Syndication
Why it matters: A hospital CISO's case for steady, standards-based PQC migration.
John Frushour, CISO of New York-Presbyterian, says healthcare should prepare for post-quantum cryptography with standards and products already available rather than fear-driven forecasts, and that commercial software and service providers will carry much of the transition.
Cybersecurity & Threats
[P2] Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident — The Hacker News
Why it matters: A breach at an AI inference vendor cascades into a poisoned coding-agent extension.
Cursor says an attacker used an access key it had issued to inference provider Fireworks AI, exposed in a breach at Fireworks, to copy coding-agent traces from sessions outside Privacy Mode; credentials left in those traces led into Cursor's SaaS tools, and on 29 October a stolen credential let the attacker plant trojanised builds of Anthropic's Claude Code extension, with an SSH backdoor, in the marketplace cache for about 11 hours, reaching at most 5,000 users. Uber separately says drivers' licence data was stolen in October through a service provider's compromised credentials.
severity high · exploited in the wild · EU: CRA, NIS2, AI Act
[P2] Data Breach at La Poste Subsidiary Maileva Affects 22 Million People in France — Security Affairs
Why it matters: One of France's largest breaches exposes 22 million people who never dealt with the company.
Maileva, a Docaposte mailing subsidiary of La Poste, was breached in mid-September, exposing names and postal addresses of 22 million recipients of documents its 70,000 business users sent, including social-housing tenants and patients of Inovie laboratories and, through a billing subcontractor whose reminders were read, Foch Hospital; La Poste says no bank details were taken. Maileva notified the CNIL and filed a criminal complaint.
severity high · exploited in the wild · EU: GDPR, GDPR Art. 9, NIS2
[P2] Extortion Group Claims Theft of Nearly One Billion Expedia Traveler Records — Security Affairs
Why it matters: A stolen vendor key opens a path into a travel giant's data platform.
Expedia confirmed attackers reached its systems through a compromised third-party data platform provider in a supply-chain attack that hit several companies, but not the figures claimed by the new Russian extortion group n0n: over 830 million loyalty-programme records, 58 million traveller profiles and 333 million account records with password-like tokens. Wiz and Sentra say the intrusion used a stolen vendor key from Qubole, a Google Cloud partner.
severity high · exploited in the wild · EU: GDPR, NIS2 · actor n0n (Russian-speaking extortion group) (60%)
[P2] New worm-like 'SCKit' malware targets MSPs in ScreenConnect attacks — BleepingComputer
Why it matters: An unauthenticated extension turns a remote-support platform into a mass-intrusion tool.
Huntress says attackers entered cloud-hosted ScreenConnect environments through RMM+, a third-party extension whose versions 1.0.31 and 1.0.32 accepted a blank shared access key, and used Run Command to install rogue clients on at least 288 endpoints in 48 organisations within 90 minutes. The SCKit toolkit then tries to spread to technicians' machines through an already patched ScreenConnect flaw (CVE-2026-84869, in KEV since September), though researchers saw no successful exploitation of it.
severity high · exploited in the wild · EPSS 0.012 · CVE-2026-84869 · EU: NIS2, CRA
[P3] New Mior and Skyfall Attacks Let Rogue Cloud Hosts Steal Secrets From AMD SEV-SNP VMs — The Hacker News
Why it matters: Confidential computing's promise to keep the cloud operator out is broken twice on AMD.
ETH Zurich's Mior attack corrupts the table that protects SEV-SNP memory during start-up, letting a host with hypervisor and BIOS control read and write a confidential VM's memory and forge its attestation (CVE-2026-0462, CVSS 4.0 5.9); Skyfall leaks VM data with hypervisor control (CVE-2026-40671, 4.9). AMD released firmware and microcode in August, with platform firmware due to server makers in November and December.
severity medium (CVSS 5.9) · CVE-2026-0462 · EU: Data Act, NIS2, EUCS