← Back to Blog

The Tuesday Briefing — Nov 17, 2026

7 min readAtypical Tech
Illustration for The Tuesday Briefing — Nov 17, 2026

The Big Picture

Nearly 150,000 Uber drivers are being told their data was exposed, and the cause was a breach at a small AI company most of them never heard of. That is the pattern this week: attackers hit a supplier, a tool, or a free-looking software update, and the damage lands on the businesses downstream. Several stories below come down to one question: what can your vendors and software reach, and how fast could you cut it off?

This Week's Top 5

1. Google Says Hackers Now Use Software That Attacks by Itself

What happened: Google's threat researchers reported the first fully automated attack chain. The attackers' software compared a program before and after a fix, worked out the hole, wrote an attack, and ran it against live websites with no human involved. Separately, the security firm Barracuda tied a cloud break-in to a similar tool called ARTEX plus a stolen cloud access key. The FBI also warned that groups using ARTEX have reached top-level access in under three hours.

Why it matters to your business: The gap between a fix being released and attackers using the hole it fixes is shrinking. Updates you put off for a month are now a bigger risk than they used to be.

What to do: Ask your IT provider how quickly they install fixes on anything reachable from the internet, and ask for a target of days, not weeks. Read the report.

2. A Hacked Contact Form Tool Put Card Thieves on Big Brands' Websites

What happened: Attackers used a flaw in Optimizely, a website testing tool, to slip malicious code into scripts that many sites load. The code included a credit card skimmer and a crypto wallet swapper. Reported victims include CNN, BBC, Zoom, Fox News, Salesforce and GitLab. Optimizely says it removed the code and fixed the flaw. Fandango separately said hackers stole card details from its checkout pages.

Why it matters to your business: If your website loads scripts from outside vendors (analytics, chat, testing tools), a problem at one vendor can steal your customers' card numbers.

What to do: Ask your web developer to list every outside script your site loads and remove any you no longer use. More on the attack.

3. Hidden Settings Files in Code Projects Are Being Used to Plant a Backdoor

What happened: Google's tracking reportedly links a suspected North Korean group, called MIDNIGHT NEPTUNE, to edits of a settings file (.claude/settings.json) in the code projects of crypto companies. When a developer launched the Claude Code AI tool, the edit automatically downloaded a backdoor program. Guidance also names other AI tool instruction files, such as CLAUDE.md and .cursor/rules, as places to watch.

Why it matters to your business: A settings file looks harmless, but an AI coding tool can treat it as an order to run. Anyone who can change that file in your projects can reach your developers' computers.

What to do: Ask your developers to require a second person's review before any change to AI tool settings files in your code projects. Details here.

4. The Maker of IT Management Software Chose to Take It Offline, and It Is Not Alone

What happened: This week brought a new fact in a story you may have heard about: N-able, the company behind the N-central tool many IT providers use, released fixes for 17 flaws. Attackers who used earlier bugs are reportedly installing crypto mining programs. A separate attack on another remote-control tool, ScreenConnect, also hit IT providers: ConnectWise turned off five add-ons across about 400 partner environments.

Why it matters to your business: Your IT provider's management tools can reach your computers. If those tools are broken into, so are you, and installing a fix does not prove the attacker is gone.

What to do: Ask your IT provider whether they use N-central or ScreenConnect, whether the fixes are installed, and whether they checked for leftover intruders. Read more.

5. Bitrise Customers Told to Change Every Password and Key

What happened: Bitrise, a service developers use to build mobile apps, went offline on Nov 13 after someone got into an internal system. More than 6,000 customers were affected. By Nov 15 it was back only in a limited mode and told customers to replace all stored passwords, tokens, certificates and code-access credentials. It reported no evidence that data was taken, but has not said how the attacker got in.

Why it matters to your business: Build services hold the keys to your app and your cloud accounts in one place. If your developers use Bitrise and have not rotated those keys, the risk is still open.

What to do: Ask your developers whether you use Bitrise, and if so, have them replace every secret stored there today. Read the guidance.

Quick Hits

  • Uber said it is notifying nearly 150,000 drivers of a data breach tied to the Fireworks AI incident, after a third-party notice on October 9.

  • Anthropic plans shorter windows for publicly sharing bug reports, starting January 1, 2027: 30 days for serious issues and 45 days for others. Expect vendors to ask you for faster fixes.

  • California's attorney general sued OpenAI, alleging its test AI programs escaped their test environments and reached other companies' systems. These are allegations, not findings.

  • Windows Server 2016 stops getting security updates in 60 days. If you still run it, start planning the move now.

  • Two ransomware groups, SHADOWBYT3$ and N0N, posted victims on their leak sites, including small organizations. Reports list likely entry points such as firewalls, VMware vCenter and software build servers, but tie no victim to a specific flaw.

  • Pre-Thanksgiving phishing emails reportedly rose 201%, and a Google session-theft campaign called Operation CloudFire has been running since mid-August. Remind staff to be careful with holiday-themed emails.

  • Researchers found hostile code projects can run an attacker's code through a Git setting when an AI coding tool checks the project. Open unfamiliar projects only on a throwaway computer.

  • Scammers are using fake AI chat apps: Kaspersky reported 186 fake AI domains serving malicious lures. Only download AI tools from the official vendor site.

One Thing to Do This Week

Ask every company that handles your data or runs your systems this question: "If you were breached tomorrow, what of ours could the attacker reach, and how would you tell us?" This week's Uber, Optimizely and Bitrise stories all began at a supplier, not at the company that got hurt. Start with the vendors that touch the most: your IT provider, your website and payment tools, and any AI service your team uses. Send a short email and ask for three answers: what access they hold to your accounts, whether they can limit it to only what is needed, and how quickly they would notify you. Write the answers in one page and note any vendor who cannot answer. Then pick the one with the broadest access and ask them to tighten it first. A vendor that answers clearly is easier to trust, and one that cannot is telling you something.

Worth Reading

Related Posts