Secrets in build artifacts, and who gets blamed
Free weekly email
This went to subscribers on August 18. Get the next one.
One email every Tuesday: what changed in security and compliance that week, and what it means if you sell to enterprise buyers.
Free. Unsubscribe in one click, and replies reach Jacob directly. Read the latest issue or browse the archive.
Four of the five stories this week come down to the same thing: someone else's credential, sitting somewhere it should not have been, and a customer having to explain it to their customers. If you are answering security questionnaires from US buyers right now, three of these will show up as questions within a month. I have skipped the government breaches and the AI policy essays because they change nothing for you.
An AWS key in a public JavaScript bundle took down 1,000 charity CRMs
Source: SecurityWeek
CRM provider Beacon disclosed a breach affecting more than 1,000 UK charities. The suspected root cause was a compromised AWS access key that had been exposed in publicly available JavaScript build artifacts. Infosecurity put the client count above 1,500.
This is one of the most common findings I get on web application tests, and it is almost never in the source repository where the scanner is pointed. It gets baked into the built bundle, or into a sourcemap that ships to production alongside it. If your secret scanning runs on commits only, you are checking the place the key is not, and a static long-lived AWS key with any real IAM reach turns that into every record you hold.
The LiteLLM fallout is a CI credential problem, not an AI problem
Source: Help Net Security
A 153GB archive stolen in the LiteLLM supply chain attack has surfaced, containing 433,909 files. Hudson Rock attributed 118,829 CI runner dumps to 2,488 corporate domains, with credentials tied to AWS, Samsung, Cisco and Salesforce among them. The firm says it is working on a global disclosure effort.
CI runner dumps are the worst possible thing to lose, because build environments tend to hold the credentials nobody wants to rotate: cloud keys, registry tokens, signing material, production database strings. If you added an LLM gateway or proxy to your stack in the last year, treat every secret that environment could read as burned and rotate on that basis rather than waiting for a notification email. Write down when you rotated and what you rotated, because a US buyer's reviewer will ask you about this specific incident and a shrug is a finding.
Metabase SQL injection is now on the KEV list
Source: CISA
CISA added three vulnerabilities to the Known Exploited Vulnerabilities catalog based on evidence of active exploitation: a Cisco ASA and FTD heap inspection flaw, a Windows WinSock AFD use-after-free, and a SQL injection issue in Metabase. KEV entries carry remediation deadlines for US federal agencies under the binding operational directive.
Metabase is the one that matters to a 60-person SaaS company, because it is usually running on an internal subdomain, connected to a read replica of production, and owned by nobody in particular. The KEV catalogue has also quietly become the yardstick enterprise buyers use when they ask what your patching window is, so if your policy says thirty days and your Metabase instance has been untouched since spring, the policy is the liability. Go look at what that instance can query before you look at the CVE.
A year-long campaign is quietly draining Salesforce and ServiceNow tenants
Source: Dark Reading
Dark Reading reports on a campaign tracked as City-Forum, active since at least March 2025, that has been stealing data from Salesforce and ServiceNow tenants using custom tooling. Targets span multiple sectors.
Most SOC 2 scopes I read draw a boundary around the product and leave the CRM outside it, which is a fiction, because the CRM holds contract terms, support tickets with customer data pasted in, and often production credentials someone shared in a case comment. The attack path here is rarely the platform itself, it is a third-party app with a standing OAuth grant that no one has reviewed since it was installed. Pull the list of connected apps in your Salesforce org this week and see how many you can name.
When the subprocessor is breached, your customer writes the letter with your name in it
Source: HIPAA Journal
Aesto Health, a healthcare technology company based in Birmingham, Alabama, had a security incident that affected several of its healthcare provider clients. One of those clients, Quincy Valley Medical Center, has begun notifying patients and publicly explaining that the incident occurred at a third-party vendor.
If you sell software into US healthcare, this is your future press cycle, and the part worth studying is the mechanics rather than the breach. Your business associate agreements set a clock on notification, and meeting it requires knowing, per customer, exactly what data of theirs you hold and where. Nobody assembles that map during an incident, so the ones who get through this cleanly built it a year earlier and kept it current.
The theme, if there is one, is that the credential you forgot about is the one that ends up in somebody's disclosure notice. I will be watching whether the LiteLLM disclosures produce named victims this week.
Jacob
Free weekly email
Get the next issue on Tuesday
One email a week from Jacob Masse: the security and compliance stories that changed something that week, and what each one means if you sell software to enterprise buyers. Five stories, a take on each, five minutes to read.
Free. Unsubscribe in one click, and replies reach Jacob directly.