Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Package registries have a well-known abuse pattern: attackers upload malicious packages, and unsuspecting developers install them. Our researchers just found the pattern working in reverse, in a RubyGems supply chain attack that turns the registry into a place to stash stolen data rather than deliver it.

A real intrusion, captured in full, shows why trust cannot live inside the model that does the work.

Attestation is a security process that enables one system or entity to prove its state or characteristics to another. This typically involves generating verifiable evidence about the software, hardware, or configuration of a device or environment. The primary goal is to ensure that systems are operating as expected and have not been tampered with. Attestation is important for building trust in distributed environments, where direct oversight and control are not always possible.

Two years ago, your teams shipped software. Today they ship two different things. They ship software that AI mostly wrote. And they ship AI systems they built themselves: models, agents, features that reason and act. Most security programs are still scoped for the first and blind to the second. That gap is not a tooling problem. It is a category problem. And the way the industry is drawing the categories is making it worse.

The economics of continuous security at frontier-model prices, and why the math points back to independence. The frontier models are astonishing at finding vulnerabilities. That is not in dispute, and it is not what this piece is about. The question is not whether a frontier model can find a flaw in your code. It is whether you can afford to run one as your scanner, continuously, across your entire estate, the way real security actually works.

An attacker republished more than 140 packages in the @mastra npm scope, each carrying a single malicious dependency, easy-day-js. The malicious versions were observed on 2026-06-17. easy-day-js is a typosquat of the dayjs date library: version 1.11.21 is the clean prior release with no install hook, while version 1.11.22 adds an obfuscated postinstall dropper.

AI models that generate code are also the best at exploiting it. Here’s why independent verification, not the model itself, is the only trustworthy answer. This month, the US government ordered Anthropic to suspend access to its most capable models, Mythos 5 and the newly released Fable 5, for all foreign nationals, citing national security. The trigger was a single reported jailbreak that let one of those models slip past its own guardrails on cybersecurity tasks.

There was a time when “AI red teaming” sounded like a novelty. Now, it’s fast becoming table stakes. If your organization is shipping machine learning or LLM-powered systems into the real world (especially in sensitive domains), you need to know how those systems behave under pressure. That’s where AI red teaming tools come in. These tools help teams stress-test AI the way it will actually be used (and misused).

Software Composition Analysis (SCA) services are automated tools that scan codebases to find, identify, and manage open-source components, detecting security vulnerabilities (CVEs), licensing issues, and outdated libraries. They help teams maintain secure, compliant software by creating a software bill of materials (SBOM) and shifting security left in the development lifecycle (DevSecOps). Top providers include Mend.io, Snyk, and Checkmarx.

AI guardrails are safety, security, and governance frameworks designed to ensure Large Language Models (LLMs) and generative AI applications produce trustworthy, accurate, and appropriate content. They function as filters for inputs and outputs to prevent harmful or biased outputs and proprietary data leakage, enforcing compliance with safety policies and regulatory standards.