Our weekly review covers 9 to 15 November, read by a risk owner rather than an analyst. Five angles this week: threat intelligence, attack surface, compliance, risk management and AI.
The thread of the week fits in one sentence: neither the number of victims nor the score of a flaw measures your risk. A ransomware group inflated its claims until it topped the monthly rankings; a flaw rated "moderate" became a federal emergency on the day it was published. Our position: an executive team that receives counts and scores receives noise; it needs known exploitation, real exposure and a decision. It is a position, not a certainty: counts remain a useful signal. But a signal is not a measurement.
Part 1: threat intelligence, the ranking misleads
What happened
- The count. Bitdefender recorded 1,072 claimed ransomware victims in October, on the groups' leak sites. Status: reported by the researchers (Bitdefender, 10 Nov.).
- The inflation. The N0n group tops the month with 126 claimed victims. According to Bitdefender, the server the group exposed shows about twenty real victims; the rest are duplicates and inflated claims. Without N0n, the month's total falls to 946. Status: reported by the researchers; the claims themselves are claimed by the attacker.
- The pace. More than 17 new groups appeared in two months, again according to Bitdefender. The market is fragmenting, and every new group has an interest in looking bigger than it is.
What it changes
A claim on a leak site is an attacker's fact, not an incident fact. Two consequences for an executive team. First, your crisis plan must say who verifies a claim made in your name, and within how many hours, before any public statement. Second, your sector's rank in a monthly report is not your risk: what matters is whether the groups active among your peers exploit what you have exposed.
Part 2: attack surface and active exploitation
- Microsoft Outlook (CVE-2026-124194). Patch Tuesday on 10 November fixed an Outlook flaw already exploited: opening a crafted file makes Outlook contact an attacker's server without warning, handing over the user's Windows authentication hash. Rated "moderate" (6.5), it still entered the KEV catalogue the same day, with a federal deadline of 24 November. The detail that matters: the fix ships through Office updates, not Windows updates. Status: exploitation confirmed by Microsoft and CISA.
- Exposed devices. CISA added four flaws to the catalogue on 9 November, six on the 10th and two on the 13th. Among them, two Barracuda products, to fix within only three days: the catalogue's shortest deadline is reserved for what is exposed and actively attacked. Status: confirmed by CISA.
What it changes: a "moderate" rating says nothing about urgency. If your patching sorts by score, this flaw waits; if it sorts by known exploitation and exposure, it goes first. And the prior question stays the same: do you know, this week, which devices on your perimeter answer from the internet?
Part 3: compliance, the deadline that did not come
- United States, CMMC. 10 November 2026 was meant to make third-party certification the norm in US defence contracts involving controlled unclassified information: phase II of the CMMC programme. The date passed without it. The US department suspended it in July for a review; phase I, the self-assessment, is still checked before a contract is awarded. Status: announced by the department, reported by several law firms.
- Canada, CPCSC. On the Canadian side, Level 1 of the Canadian Program for Cyber Security Certification (13 controls, an annual self-assessment) is required in select defence contracts from summer 2026. Status: confirmed by Public Services and Procurement Canada, which announced it in April.
What it changes: a compliance calendar moves, the required controls do not. A Canadian supplier in both defence supply chains answers to both regimes; a suspended phase is not an exemption, and it is evidence of the controls that will be requested, whatever the date.
Part 4: risk management, in Canada
The Canadian Investment Regulatory Organization (CIRO) proposes to pay $8.3 million to settle the class action born of the breach detected in August 2025, according to Advisor.ca (10 Nov.). About $4.4 million would go to class members (extended credit monitoring, documented losses up to $10,000 per person), the rest to administration, notice and legal fees. The class counts more than one million people; the approval hearing before the Superior Court of Quebec is scheduled for 14 January. CIRO denies any liability. Status: reported, based on the proposed settlement.
Further north, Air Creebec said on 6 November that unauthorized parties accessed some of its administrative systems in late October and took files that may contain sensitive information; the number of people affected is not yet known (reported by CTV). Elsewhere, TSMC told the SEC on 12 November that employee data had leaked through a cloud service provider, with no sign that its own systems were compromised (confirmed by the company).
What it changes: the CIRO figure is useful to a board because it lands fifteen months after the incident, on top of the response itself. A register line that prices a breach at the cost of the response forgets the tail: notices, credit monitoring, class action. In Quebec, that tail is also argued in court.
Part 5: AI on both sides of the table
Attack side. Bitdefender reports that N0n spent more than $7,000 on generative AI services to run its operations, and that an affiliate of another group connected an AI coding assistant to a remote shell on its victim's machine. According to TechCrunch, the attackers of Fireworks AI mainly want to resell access to large language models. Status: reported.
Defence side. On 9 November, Notion notified customers that Notion AI conversations used for its Insights feature were reached at Fireworks AI, the provider running its models, for fewer than 1% of workspaces. Notion advises treating the data as exposed and rotating any password or API key pasted into Notion AI during the period. Status: confirmed by Notion.
What it changes: AI lowers the cost of an attack for groups that are still small, and every AI tool you adopt adds a provider that sees your prompts. What your employees type into an AI tool is data to classify, just like a file.
Three questions for an executive committee this week
- If a group claims an attack in our name tomorrow, who verifies it, within how many hours, and who speaks?
- Does our patching order follow known exploitation and exposure, or only the flaw's score?
- Which AI tools receive our prompts, and are they in our register as third parties?
What do you think?
Our position: counts and scores are signals, not measurements of risk. Does your executive team receive numbers or decisions? Tell us in the comments what your cyber dashboard contains.
Forta Radar sorts the threat by sector and by known exploitation, and the Exposure module shows what answers from the internet on your side. Our free exposure report delivers it within 48 hours, and thirty minutes with an expert is enough to work through your three questions.
Sources: Bitdefender, Threat Debrief, November 2026 (10 Nov. 2026) · CISA, additions to the KEV catalogue (10 Nov. 2026) · CISA, KEV catalogue · Public Services and Procurement Canada, CPCSC Level 1 · CIRO, cybersecurity incident notice · TSMC, Form 6-K filed with the SEC (12 Nov. 2026) · Notion, security advisory on an incident at a third-party AI vendor (9 Nov. 2026). Monitoring, named in the text: TechCrunch, Advisor.ca, CTV, law firms on CMMC.