Fairlife's Milk Machines Are Offline

Coca-Cola filed an SEC disclosure this week confirming that Fairlife, its ultra-filtered milk and protein shake brand, was hit by a ransomware attack that has suspended all US production, according to Bleeping Computer.

Attackers gained unauthorised access to Fairlife's systems, including the production infrastructure that keeps the factory lines running. Once ransomware encrypts those systems, operations grind to a halt until they are restored or rebuilt. Canadian facilities are still running, but US output of Core Power protein shakes, Nutrition Plan drinks, and Ultra-Filtered Milk is on ice.

No ransomware group has claimed responsibility yet. Coca-Cola has not confirmed whether data was stolen or whether an extortion demand has been received. That silence is not unusual — early-stage ransomware incidents typically follow a pattern: encrypt, steal, then threaten to publish the stolen data unless a ransom is paid.

Product safety has not been compromised. The business impact, however, is still being calculated.

What you should do: If you rely on Fairlife products professionally, plan for short-term supply disruption. More broadly, this is a good reminder that ransomware hits physical production, not just data. If your organisation has operational technology connected to the internet, that is a risk worth reviewing.

Researchers from Seoul National University, the University of Illinois Urbana-Champaign, and Largosoft have published a paper describing a new attack class called agent data injection (ADI), detailed by The Hacker News.

AI agents read two types of input: instructions (what you tell them to do) and data (everything they pull in while working, like a web page or email). Most defences are built to catch smuggled instructions hidden in data. ADI works differently. It corrupts the small facts the agent quietly trusts — a sender's name, a button's ID, a tool result — without ever writing anything that looks like an order.

In tested attacks, a planted product review made a web agent click "Buy Now" instead of "Read More." A forged GitHub comment made a coding assistant run a stranger's command on a developer's machine.

What you should do: Be cautious about approving AI agent actions without reviewing exactly what they are about to do. Treat every agent confirmation prompt the way you would treat an unfamiliar permission request — read it before you click yes.

What FortiSandbox is: A Fortinet security product that analyses suspicious files and network traffic in an isolated environment to detect malware before it reaches the rest of a network.

What it is: An OS command injection vulnerability (where an attacker sends specially crafted input that the system mistakenly runs as a system-level command) that allows an unauthenticated attacker to execute arbitrary commands via malicious HTTP requests.

Who's at risk: Any organisation running an unpatched FortiSandbox deployment reachable from the internet. No login required to exploit this.

CVSS: 9.1 — Critical. Patch today.

Root cause: The software fails to properly sanitise user-supplied input before passing it to underlying system functions. Think of it like a form that accepts your name but also accidentally runs anything you type as a terminal command.

Attack vector: An attacker sends a crafted HTTP request to the FortiSandbox interface — no credentials, no user interaction needed. The system processes the malicious input as a command and executes it with elevated privileges.

Recommended actions:

1. Upgrade to FortiSandbox 4.4.9 or 5.0.6 immediately.
2. If running cloud or PaaS versions without an available patch, discontinue use until patched.
3. Restrict external access to the FortiSandbox management interface at the network level.

1Password has released a beta integration with Claude that lets the AI assistant complete browser tasks requiring a login — without ever seeing your actual password, according to Help Net Security.

Here is how it works: when Claude hits a sign-in page during a task, it requests the credential from 1Password. You approve the request using Touch ID or another biometric method. The password is filled in through a secure channel that never enters Claude's memory or Anthropic's systems. Once the task ends, the permission resets.

This is available to paid Claude subscribers (Pro, Max, Team, or Enterprise) on macOS with the 1Password desktop app and browser extension installed.

If you use Claude for browser-based tasks, this is worth enabling. It keeps your credentials out of AI context windows entirely — which is exactly where they should not be.

CISA added two critical Fortinet FortiSandbox vulnerabilities — CVE-2026-39808 and CVE-2026-25089, both rated CVSS 9.1 — to its Known Exploited Vulnerabilities catalogue on July 16, according to Infosecurity Magazine. US federal agencies must patch by July 19 under Binding Operational Directive (BOD) 26-04. For cloud deployments without available patches, CISA says to stop using the product entirely. Private sector organisations are not legally bound by BOD deadlines, but with active exploitation confirmed, the same urgency applies.