That's the harsh reality for those of us who work in cybersecurity. 💪 > Assume exposed software is brittle. > The stuff is broken. Plan accordingly. 👇

Universal Memory Protocol universalmemoryprotocol.io/ A transport-neutral memory protocol for AI agents. What MCP did for tools, UMP does for memory - negotiated operations over a portable, signed, bi-temporal record that any harness can speak and any store can serve.

⚡"Breaking Post Quantum Cryptography with AI" A non-profiled deep-learning side-channel attack on an unprotected reference implementation. The convolutionnal neural network just plays the role CPA's correlation used to play. The @DonjonLedger 's PQC journey continues. They pointed their open-source deep-learning SCA tooling at the NIST-standardized ML-KEM reference. No clone device. No profiling phase. No fixed leakage model. Only EM traces, chosen ciphertexts, and a small MLP trained per key hypothesis. The correct key is the one under which the network actually learns. ~400 traces. Unprotected target, no masking, no shuffling. - ML-KEM is mathematically sound and standardized. - A reference implementation running on a real chip, without countermeasures, leaks the secret in minutes. PQC security does not stop at standardization. It starts when implementations meet real-world attackers, with probes, not just headlines. Read the article: donjon.ledger.com/blog/non-p…

This is really big news. Google introduced the Open Knowledge Format (OKF) - a standardized way to store information in a directory of markdown files. Makes it really easy to make a digital brain that agents can use. These files can serve as a living wiki. You can give agents the ability to query them or edit them. They can interlink. Seems to me this could replace Notion or Obsidian. I can think of so many uses for this. Google's blog post: cloud.google.com/blog/produc… An easier to understand explanation is the SPEC.md file: github.com/GoogleCloudPlatfo… I gave those two links to Antigravity and asked how we could use it for any of the projects we're working on. It came up with so many ideas. I would imagine Claude Fable 5 would whip up some pretty amazing things based on this system. Currently creating an OKF library of our pepper garden. It's going to be a fun weekend.

‼️ Anthropic's recently released frontier model Fable 5 was jailbroken by someone using a jailbroken version of Claude Opus. The researcher who goes by the moniker pliny carried out the jailbreak and says: "the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement" The jailbroken version can be used for research into and exploitation of vulnerabilities.

I don’t care if “our enemies” get access to some super hacker AI. Just assume they have it and plan accordingly. I’m far more worried about AI being controlled by a small number of trillion dollar companies that decide who gets access and what you’re allowed to do.

NIST has a useful paper on AI guardrails The takeaway is that static guardrails are the wrong security model for open-ended LLM systems. A finite set of rules cannot cover every adaptive prompt. You can harden the system, make bypasses harder, monitor for abuse and reduce the blast radius. But you should not patch an LLM once, add a few refusal rules and call it done. LLM security needs to look more like vuln research and detection engineering: continuous testing, continuous updates and an assumption that bypasses will eventually be found nist.gov/news-events/news/20…

Yacine is right and we’re all going to have to save up for massive home compute clusters now if we want to fix this

As predicted, they’re hitting rust crates now. It’s not perfect or prod-ready yet but sanctum helps to mitigate this: github.com/postrv/sanctum-os…

This is the kind of Linux bug attackers love. > Not remote > Not flashy > Easy to miss But once they already have a small foothold, it can turn that access into full control of the host. CVE-2026-23111 now has public exploit details. Patch reboot: thehackernews.com/2026/06/on…

The reason is quite hilarious 😂😂. Microsoft put $50 billion into Anthropic. FIFTY billion dollars. they are a Project Glasswing partner. Fable 5 runs inside Azure. Microsoft sells Claude to its own enterprise customers through Microsoft 365 and GitHub Copilot. and they won't let their own employees use it. here's why. under Anthropic's new Mythos-class data retention policy, every prompt you type and every response you get is stored for 30 days. automatically. no opt out. if their safety classifiers flag anything in your session, anything, they keep it for up to two years. you don't get told when that happens, what was flagged or who can see it. Microsoft employees paste confidential contracts into these things. customer data. internal roadmaps. acquisition strategies. legal documents. source code. all of it sitting on Anthropic's servers for 30 days minimum. flagged sessions for two years. so the company that invested $50 billion looked at that policy and told its staff: actually hold on. other Claude models still work internally. under Zero Data Retention rules. the normal ones are fine. just not the most powerful one they helped fund. and one more thing. the Pentagon listed Anthropic as a supply chain risk in March and banned defense contractors from using its products. Microsoft funds Anthropic. sells Anthropic's models. runs them on Azure. helped build the most powerful one. won't let employees use it. the Pentagon won't let defense contractors near it. the safeguard that makes Fable 5 safe enough to release publicly is the same safeguard that lets Anthropic keep your data for two years. the guardrail is a data retention policy. but you can use it. it's in your browser right now. 🌚 have fun.

🏝️ OASIS — AI-Powered Code Security Auditing with Ollama Scan codebases for security vulnerabilities using local LLMs through Ollama. OASIS combines multi-model analysis, LangGraph orchestration, vulnerability validation, rich reporting, and an interactive web dashboard—all while keeping your source code on your own infrastructure. 🔗 github.com/psyray/oasis #CyberSecurity #AppSec #CodeReview #DevSecOps #Ollama #LLM #SecureCoding #OpenSource #SecurityAudit #AI

If you use @browserling for cross-browser testing, you can also add a lightweight security testing layer with OWASP PTK. Watch the demo.

We have detected that the popular package `onering` on crates.io has been compromised with an information stealer that runs on build, which sends a git diff to a Sentry endpoint without authorization: github.com/cenotelie/onering… This is quite novel.

⚡ Microsoft is adding a 2-hour delay before VS Code extensions auto-update. The wait gives maintainers more time to catch bad or compromised releases before they spread further. Read details: thehackernews.com/2026/06/vs…

.@trailofbits holds the world record for breaking elliptic curve crypto with the fewest quantum qubits. Woooo 🤘

yv1ing/z3r0: A controlled multi-agent workbench for authorized security assessment, code auditing, internal review, and controlled research. github.com/yv1ing/Z3r0

Whisper Leak — Exposing Side-Channel Information Leakage in Streaming LLM Conversations 🤖💀 Whisper Leak is a security research toolkit that demonstrates how encrypted, streaming conversations with Large Language Models can leak prompt information through packet sizes and timing patterns, even without decrypting TLS traffic. Key Features: • Capture and analyze encrypted LLM traffic • Build datasets from real chatbot sessions • Train CNN, LSTM, BERT, and LightGBM classifiers • Benchmark multiple LLM providers and configurations • Perform live or PCAP-based inference • Evaluate side-channel leakage risks at scale • End-to-end workflow for collection, training, and testing Supported across major LLMs including GPT, Claude, Gemini, DeepSeek, Grok, Qwen, Llama, Mistral, and more. 🔗 github.com/yo-yo-yo-jbo/whis… #CyberSecurity #AIsecurity #LLM #ThreatResearch #OpenSource

Nullsec-S1 — Security-Focused AI Model for Auditing AI-Generated Applications 🤖💀 Nullsec-S1 is an open-source security model built specifically to audit AI-generated apps, AI agents, MCP tools, Web3 workflows, and vibecoded software. Unlike general-purpose LLMs, Nullsec-S1 produces structured JSON security audits containing findings, severity, exploit scenarios, remediation guidance, secure patches, and deterministic safety decisions. Key Highlights: • Built on Qwen2.5-Coder-7B-Instruct using PEFT/QLoRA • Covers 16 security categories including SSRF, SQLi, XSS, Prompt Injection, MCP Tool Abuse, Command Injection, and Web3 risks • Deterministic Security Alignment & Safety Layers • MCP support for Cursor and Claude Desktop • CLI, API, FastAPI server, and CI integration • 1,741 curated security training examples • 111-case security benchmark focused on AI-generated software Benchmark Results: 🥇 Nullsec-S1 — F1: 0.9245 | False-Safe Rate: 0.0% 🥈 GPT-5.3 Codex — F1: 0.7252 🥉 Claude Opus 4.8 — F1: 0.6550 The project focuses on structured, reproducible security verdicts rather than free-form security commentary. 🔗 github.com/trynullsec/nullse… #CyberSecurity #AISecurity #AppSec #MCP #OpenSource

Microsoft discovered that Anthropic's Claude Code GitHub Action is vulnerable to prompt injection attacks via issues and Pull Requests microsoft.com/en-us/security…

I saw the Gemma 4 announcement two days ago, found some time tonight. Built a small CLI around it. Gemma 4 12B can describe images, and it's just an 8GB model you can self-host, which means you can wire it into small utility tools without paying for a vision API. I used it to generate alt text & blog outlines from whiteboard photos Works well enough to be useful :D