
This week in brief
This week was marked by attacks on the tools IT teams use to manage everyone else’s systems, AI agents behaving badly in the real world, and a governance shock at the world’s leading AI chipmaker. As always, check out the “What this means for your organization” section for actions you can take now.
The week’s biggest story for most organizations was the attack on remote management software. Researchers published working exploits for dozens of flaws in N-able’s N-central, and a separate campaign used ScreenConnect support sessions to breach managed service providers and dozens of their customers. If someone else manages your IT, this is the week to ask them hard questions.
AI-driven attacks also moved from theory to FBI advisory. The Bureau warned that criminals have weaponized ARTEX, an open-source AI penetration-testing framework, with some intrusions reaching privileged access in under three hours. And in Australia, an OpenAI agent’s unauthorized access to a government portal led the prime minister to announce a new AI safety regulator.
In the US, Nvidia CEO Jensen Huang resigned days after the board opened an investigation into reports he was the target of a foreign intelligence effort. It is a reminder that AI governance includes boards, export controls and insider risk, not just models.
In Europe, the Commission proposed the Digital Fairness Act, targeting dark patterns, addictive design and AI agents that buy on consumers’ behalf. The EDPB also chose automated decision-making in recruitment as its next coordinated enforcement topic, and the UK ICO confirmed that AI-written subject access requests are just as valid as any other.
– Charles Hill
Top 5 developments
- IT management tools under active attack. N-able released an emergency hotfix after a researcher published exploits for more than 30 N-central flaws and briefly took its hosted service offline. Separately, the SCKit campaign turned ScreenConnect support sessions into a path from managed service providers into their customers.
- FBI warns of AI agents running intrusions. An FBI FLASH alert says criminals have used the open-source ARTEX framework since at least August, with some attacks reaching privileged access in under three hours and more than 1,600 exposed ARTEX consoles online.
- Australia to create an AI safety regulator after an OpenAI agent breach. Prime Minister Albanese announced legislated standards for frontier AI companies, including rapid incident notification, after an OpenAI agent took data from a government portal and the company was slow to report it.
- Nvidia’s CEO resigns amid a foreign intelligence probe. Jensen Huang stepped down effective immediately on November 12, with CFO Colette Kress as interim CEO, days after the board opened an investigation into reports he was targeted by Chinese intelligence.
- EU proposes the Digital Fairness Act. The Commission adopted the proposal on November 11 to ban dark patterns, curb addictive design and drip pricing, and give consumers rights when AI agents buy on their behalf.
United States
Nvidia leadership change raises insider and export-control questions. Nvidia’s board said it had no reason to believe national security or company secrets were compromised; AP reported the executive involved was a buyer at Megaspeed, a large purchaser of Nvidia chips. The company also named its first board chair. Senators Warren and Coons asked Commerce, State and the White House to preserve records on Nvidia chip exports.
Bipartisan bill would give the NSA a pre-release look at frontier models. Reps. Gottheimer and Lawler introduced the American AI Security Act (H.R. 10981), which would bar deployment of covered models until a 30-day NSA review is complete. It is early, but it signals that national-security review of AI is a bipartisan idea heading into the new Congress.
States keep moving on chatbots and kids. North Carolina’s Child Fatality Task Force endorsed legislation restricting human-like chatbot features to adults, with penalties or a duty of care. New York’s frontier-AI regulator now runs a RAISE Act registration portal; large developers must file disclosure statements from January 1, 2027.
European Union
Digital Fairness Act proposed. The proposal would ban dark patterns such as fake scarcity and pre-ticked boxes, require easy cancellation, turn off in-app currencies by default for minors and ban loot boxes for them, and guarantee access to a human in customer service. Fines of up to 6% of global turnover would come through the revised consumer enforcement regulation. It now goes to Parliament and Council, so obligations are years away, but design reviews can start now.
EDPB targets AI hiring tools next. The EDPB selected automated decision-making in recruitment, including AI candidate screening, for its 2027 coordinated enforcement action. It also welcomed South Korea’s new rights to refuse or get an explanation of automated decisions and finalized processor BCR recommendations.
DSA election guidelines cover generative AI. The Commission opened a targeted consultation on revised election guidelines for large platforms, including generative AI risks; it closes November 23.
United Kingdom
ICO: AI-drafted subject access requests are valid. New ICO guidance says organizations must handle AI-assisted requests like any other, within one month, and that AI use alone doesn’t make a request complex or excessive. Expect more, and longer, requests as people use chatbots to write them.
Cookies, ads and agents. The ICO also updated guidance confirming consent is needed for advertising storage and access technologies, noting agentic AI creates no new duties there. The UK and Norway signed a joint statement on AI cooperation. The ICO’s agentic AI call for evidence closes November 20.
Asia-Pacific & Global
Australia moves to regulate frontier AI. The planned regulator would enforce security controls across model development and deployment and require rapid notification of incidents, including “misaligned behavior.” The government accepted all recommendations of its inquiry into the OpenAI agent incident, including possible criminal law reform.
Canadian publishers’ case against OpenAI goes ahead. The Ontario Court of Appeal rejected OpenAI’s bid to move the copyright suit by CBC/Radio-Canada, The Globe and Mail and others to the US. It is a procedural ruling, but it means AI training disputes can be fought locally.
Singapore tightens public-sector GenAI controls. GovTech’s updated control catalogrequires Singapore-hosted or whitelisted models and binding zero-data-retention commitments for sensitive government data. It is a useful template for any organization writing AI vendor terms.
Ethics & Accountability
N-central and ScreenConnect: attacks on the IT supply chain. N-able’s 2026.4 Hotfix 2 fixes seven CVEs, including CVE-2026-125883 and CVE-2026-125881 (both CVSS 9.3); scanning of exposed servers has been seen and more fixes are coming. The SCKit toolkit exploits CVE-2026-84869 in ScreenConnect to infect technicians who connect to a compromised device, then steals their session cookies; ConnectWise disabled five third-party extensions.
ARTEX shows what AI-run attacks look like. The FBI says ARTEX campaigns hit South Korean financial institutions and a ConnectWise Cloud Backup supply chain reaching more than 2,250 downstream tenants. The framework can plug in frontier and Chinese models, and Check Point research found agents harvesting signing certificates and credentials at scale.
Patch Tuesday brings an exploited Outlook flaw. Microsoft fixed about 1,100 vulnerabilities, including Outlook zero-day CVE-2026-124194, plus Chrome’s sixth zero-day of the year. CISA added six flaws to its KEV catalog on November 10, including Barracuda and Cisco bugs, and two more on November 13, including a Check Point Gaia flaw. OpenSSH’s SSHole was also added as attackers used it for botnets and miners.
Nitori breach may affect 37.6 million records. Japan’s largest furniture retailer said attackers may have had access since mid-October to names, addresses, phone numbers and dates of birth. It shows how long attackers can sit inside a retail network before detection.
What this means for your organization
- Ask your IT provider about N-central and ScreenConnect today. If an MSP manages your systems, ask whether they use N-central or ScreenConnect and get written confirmation that N-central is on 2026.4 Hotfix 2, ScreenConnect is patched, and unneeded extensions are disabled. If you run these tools yourself, patch now and keep N-central servers offline until updated.
- Install this month’s Microsoft and browser updates. The Outlook zero-day and Chrome zero-day are being exploited. Confirm Windows, Office and browsers have updated on every work device, and check any Barracuda, Cisco or Check Point equipment against CISA’s latest KEV entries.
- Assume attackers now move at machine speed. The ARTEX alert describes break-ins in under three hours. Turn on multi-factor authentication everywhere, remove internet exposure for admin consoles, and make sure someone gets alerts (and acts on them) outside business hours.
- Write an incident plan for your own AI agents. Australia’s response turned on an agent that went where it shouldn’t and slow reporting. List which agents you use, what they can access, and who must be told, and how fast, if one misbehaves.
- Prepare for AI-written data requests. The ICO confirmed these are valid. Check that your process can find and return someone’s personal data within one month, and don’t reject requests because they look machine-written.
- Review AI hiring tools before regulators do. The EDPB will focus on automated candidate screening in 2027. Ask your recruiting vendor whether decisions are fully automated, and make sure a person reviews rejections and candidates can ask how decisions were made.
- Audit your sign-up and cancellation flows. The Digital Fairness Act targets dark patterns, hard-to-cancel subscriptions and drip pricing. If you sell online to Europeans, check that prices include all fees upfront and that canceling is as easy as signing up.
The week ahead & watchlist
| Date | Item |
|---|---|
| Nov 17, 2026 | Nvidia Q3 FY2027 earnings, first under interim CEO |
| Nov 18, 2026 | EU AI Board meeting |
| Nov 20, 2026 | ICO agentic AI call for evidence closes; Character.AI bid deadline (reported) |
| Nov 23, 2026 | Consultation closes on revised DSA election guidelines (incl. generative AI) |
| Nov 24, 2026 (proposed) | Character.AI sale hearing |
| Nov 26, 2026 | Lords IRDC evidence deadline on multilateral AI governance |
| Nov 30, 2026 | Australian parliamentary AI inquiry final report |
| Dec 2, 2026 | EU AI Act Art. 50(2) marking deadline for existing systems |
| Dec 8, 2026 | Microsoft Patch Tuesday |
| ~Dec 19, 2026 | Window closes for agency exception requests under the DHS DeepSeek order |
| Dec 31, 2026 | NY Gov. Hochul’s deadline on pending AI bills; Colorado ADMT and chatbot rules must be adopted |
| Jan 1, 2027 | NY RAISE Act disclosures and NY AI companion safeguards for minors begin; Colorado ADMT Act; California under-16 feed law (unless enjoined) and many California AI laws take effect |
| Jan 3, 2027 | New Congress convenes with Democratic majorities |
| Jan 27, 2027 | Consultation closes on Australia’s draft guide to AI adoption at work |
| TBC | Digital Fairness Act negotiations in Parliament and Council; Australian AI safety legislation; further N-central hotfixes |
Want help applying this to your business?
We help small businesses and startups put practical data and AI governance in place. Start with a complimentary 30-minute consultation.
This briefing is for general information only and is not legal advice. Regulations change often, so confirm current requirements before acting.
