Data Safety and Information Security

In 2026, hardly any smartphone brands are immune to the risk of the market passing them by.

AI is rapidly changing how software is written, deployed, and used. Trends point to a future where AIs can write custom software quickly and easily: “instant software.” Taken to an extreme, it might become easier for a user to have an AI write an application on demand—a spreadsheet, for example—and delete it when you’re done using it than to buy one commercially. Future systems could include a mix: both traditional long-term software and ephemeral instant software that is constantly being written, deployed, modified, and deleted. AI is changing cybersecurity as well. In particular, AI systems are getting better at finding and patching vulnerabilities in code. This has implications for both attackers and defenders, depending on the ways this and related technologies improve...

A Conceptual Field Topology for Proactive Cognitive Defense

The European Union's Cybersecurity Service (CERT-EU) has attributed the European Commission cloud hack to the TeamPCP threat group, saying the resulting breach exposed the data of at least 29 other Union entities.

Why npm, pip, and direct Git dependencies can expose your secrets, how the attack works, and which controls actually reduce the blast radius.

Hijacked maintainer account used to publish poisoned axios releases including 1.14.1 and 0.30.4. The attacker injected a hidden dependency that drops a cross platform RAT. We are actively investigating and will update this post with a full technical analysis.

Chrome's DevTools Protocol leaks its own presence through V8's object serialization. A classic CDP detection signal was patched — this one wasn't.

Agents don't just perform actions; they execute plans where the safety of each step depends on what has already happened. That makes sequencing an authorization problem. This post explores how policy, delegation data, and multi-signature approval can govern the order in which agents receive authority, not just the scope of it.

Cross-domain delegation requires more than transferring a credential. In a society of agents, policies define boundaries, promises communicate intent derived from those policies, credentials carry delegated authority, and reputation allows trust to emerge through repeated interactions.

An in-depth investigative report on the March 2026 LiteLLM supply chain attack. Discover how the Trivy GitHub Actions hack led to a massive PyPI compromise

Analysis of malicious telnyx 4.87.1 and 4.87.2 on PyPI — a package with over 1 million monthly downloads: injected code uses WAV audio steganography to deliver payloads that steal credentials and establish persistence. Attributed to TeamPCP.

MAESTRO (Multi-Agent Environment, Security, Threat, Risk, & Outcome) is a novel threat modeling framework for Agentic AI. Assess risks across the AI lifecycle.

Cryptography is a funny thing. Supposedly, if you do the right kind of maths to a message, you can send it off to somebody else, and as long as they’re the only one that knows a secret little…

Three Claude.ai vulnerabilities chained into a full attack: prompt injection to silent data exfiltration. Oasis Security research disclosure.

Unicode that's invisible to the human eye was largely abandoned—until attackers took notice.

archived 15 Mar 2026 15:46:35 UTC

A web scraper to download the Apple Platform Security Guide as Markdown for use in Obisidian and elsewhere.

This is long compilation of all the recorded MCP security flaws in the wild.

I tried six different approaches to protect my credentials from an AI coding agent. Encrypted files, keychain, ACLs, Touch ID, a native addon, sandboxes. Every one failed for the same reason.

An overview of storage-encryption options in Linux.

Attacks against modern generative artificial intelligence (AI) large language models (LLMs) pose a real threat. Yet discussions around these attacks and their potential defenses are dangerously myopic. The dominant narrative focuses on “prompt injection,” a set of techniques to embed instructions into inputs to LLM intended to perform malicious activity. This term suggests a simple, singular vulnerability. This framing obscures a more complex and dangerous reality. Attacks on LLM-based systems have evolved into a distinct class of malware execution mechanisms, which we term “promptware.” In a ...

Wasmtime is a standalone runtime for WebAssembly (Wasm), developed by the Bytecode Alliance, designed to execute WebAssembly modules

confusable-vision renders every TR39 confusable pair across 230 macOS system fonts and measures visual similarity with SSIM. 96.5% of confusables.txt is not high-risk, but 82 pairs are pixel-identical in at least one font.

This article guides you through the process of publishing solutions to Microsoft Sentinel.

Learn how to create a codeless connector in Microsoft Sentinel using the Codeless Connector Framework (CCF).

TruffleHog is an open-source secrets scanning tool that digs deep into your code to find secrets, passwords, and sensitive keys that you may have inadvertently committed.

Google spent over a decade telling developers that Google API keys (like those used in Maps, Firebase, etc.) are not secrets. But that's no longer true.

We measure the capabilities of LLMs to deanonymize users online.

Lessons from 2025 and insights for building observable AI systems in productions

Your cloud isn’t at risk because it’s fast; it’s at risk because thousands of invisible machine identities outlive your control.