Scenario Solutions Inc,

It takes more than a good idea.

Weekly Supply Chain Series: The Name on the Commit Two open source compromises on the same day. In both, everything looked legitimate. On October 8, attackers used the hijacked GitHub accounts of two well known open source maintainers to commit a workflow called "Security Audit" into 345 repositories, in two automated sweeps lasting minutes, StepSecurity reported. One was Uber's athenadriver, where the maintainer still had write access as its original author. The commit went straight to the main branch. No pull request. StepSecurity confirmed the stolen data reached the attacker's server four seconds after the workflow started. The workflow had no security function. It collected the repository's stored secrets, then searched the full git history for cloud, AI and source control credentials, including ones developers believed they'd deleted, according to Socket. The same day, a malicious version of tensorlake was published to npm from the project's own main branch, under a maintainer's identity, carrying a valid npm provenance attestation, StepSecurity found. It stole credentials on install. If the victim revoked the stolen GitHub token within 24 hours, a background job deleted their home directory. Every signal said trust this. A maintainer's name. A real repository. A provenance attestation. A "security" label. An attestation proves which repository and pipeline built a package. Whether the code in that repository was legitimate is a separate question. A maintainer's name tells you whose account was used, not who was using it. What to check this week: -Require pull request review for any change under .github/workflows. -Require approval before workflow runs. StepSecurity saw that one setting stop this campaign's exfiltration in at least one repository. -Rotate any credential ever committed to any branch, not just the secrets stored today. -Review who still has write access to repositories they no longer maintain. Which of your dependencies would you notice changing hands? Sources: StepSecurity on GhostAction: https://lnkd.in/dTfWcq4k Socket: https://lnkd.in/eHTiMj93 StepSecurity on Tensorlake: https://lnkd.in/d7SfzfK3 Cloud Security Alliance: https://lnkd.in/gtQXf46X #SupplyChainSecurity #OpenSource #GitHub #DevSecOps

The Vendor Control Surface Two supply chain breaches this year. Two different verbs. Authenticate. In a September 8 SEC filing, Veradigm, the health records company formerly known as Allscripts, said an attacker took credentials from one of its vendors' environments for a Veradigm API used for customer services, then used that access to copy patient data, including Social Security numbers for some patients, HIPAA Journal reported. Nothing else in Veradigm's network was accessed. The access was valid. Update. On March 24, two malicious versions of LiteLLM, a widely used open source AI gateway, were live on PyPI for about 40 minutes, according to LiteLLM. The company said a maintainer's publishing account was likely compromised, the attacker bypassed its official release pipeline, and the code was built to steal credentials. Anyone who installed the latest version in that window got the attacker's code through a trusted update channel. Most vendor reviews ask what a vendor says it does. A better question: what can it actually do inside your environment? Seven verbs cover most of it. Can the vendor: 1. Observe your data or activity? 2. Modify your systems or configurations? 3. Disable your services or controls? 4. Update software you run? 5. Authenticate as a user or service? 6. Decrypt anything you protect? 7. Administer anything you own? What to check this week: -List every credential a vendor holds into your environment: API keys, service accounts, support logins. Confirm each has an owner and the narrowest scope that works. -Rotate vendor held credentials on a schedule, and immediately when a vendor discloses an incident. -Pin dependencies to exact, verified versions instead of pulling whatever is latest. Which of the seven do your vendor reviews actually cover? Sources: https://lnkd.in/g7PkgYAS #SupplyChainSecurity #VendorRisk #ZeroTrust #Cybersecurity

Ignorance vs. Negligence The difference between ignorance and negligence is knowing and not knowing. Shadow AI is where that line matters most right now. In the 2026 Ponemon and DTEX insider risk study, negligent employees caused 53% of insider incidents, averaging $747,107 each, and unapproved apps, including AI tools, were the top driver. Only 18% of organizations had fully built AI governance into their insider risk programs. In May, Community Bank, a Pennsylvania lender owned by CB Financial Services, filed what's been called the first SEC Form 8-K triggered by shadow AI. An employee building a presentation put customer names, Social Security numbers and dates of birth into an unapproved AI tool on a personal account. The bank already offered an approved AI tool. The employee reportedly thought the sensitive data had been removed from the file. With CUI, it's worse. CUI pasted into a personal AI account has left its authorized boundary and may have to be handled as a spill. Ignorance is leadership's to fix. If no one was trained and no policy named the approved tools, the employee can't carry all the blame. Once training and a documented process exist, ignorance is off the table. What's left is negligence, and that's on the individual. The same process protects the people who follow it. What to check this week: -your approved AI tools and the data each one is cleared for. -Label CUI and IP so people recognize it before they paste it. -Build AI use into onboarding, annual training and offboarding. Is your team operating on knowing, or not knowing? Sources: Ponemon Institute and DTEX, 2026 Cost of Insider Risks; Infosecurity Magazine; Bloomberg, 2023 #ShadowAI #InsiderRisk #CUI #CMMC

258 days. That's IBM's 2026 average to identify and contain a breach that started with a compromised supplier. Two incidents this year show what a window that long looks like. NYC Health + Hospitals. Attackers were inside the network from November 2025 until February 2026, getting in through a breach at a vendor the health system didn't name. They copied medical records, billing data, ID documents, and fingerprint and palm print scans for at least 1.8 million people. Detected February 2. Disclosed in May. Defense Manpower Data Center. A vulnerability in a file sharing system let unauthorized users reach unencrypted personnel records from October 2025 until it was found on July 16, 2026. About 3.05 million people were affected, with Social Security numbers and military job details exposed. Notification letters went out in September. Different causes. Same pattern: access nobody was watching, for months. What to check this week: -List every vendor portal and file sharing system that holds sensitive data, and who can reach each one. -Ask your critical vendors when they last reviewed access logs for your data, and how far back those logs go. -Confirm sensitive files are encrypted at rest, not just in transit. How often does your organization revisit a vendor approval after it's signed? Sources: IBM Cost of a Data Breach Report 2026; TechCrunch, May 18, 2026; SecurityWeek and Military Times, September 2026 #SupplyChainSecurity #VendorRisk #ContinuousMonitoring #Cybersecurity

Save the Date: MACH37's Fall 2026 Launch Day! 📣   Our six innovative startups have been hard at work since joining the program in September, refining their strategies, challenging assumptions, and building toward the next stage of growth. Join us on Thursday, December 3rd at 12pm EST to hear from AnzenOT, MoodMetrics AI, Adjudica, Scenario Solutions Inc, Kognos AI, and Radar 36. Launch Day is a great opportunity to hear directly from our founders as they share the challenges their tackling, showcase the innovative solutions they're bringing to market, and highlight what's next. 🚀 We can't wait to see you there! Don't forget to save the date & register here 🔗: https://luma.com/6zh9qrwo