Tag: security

#openai #astra #pause

https://openai.com/index/responding-next-frontier-critical-cyber-capabilities/

#ki #deepfake

https://www.heise.de/news/Deepfakes-in-Videokonferenzen-Fraunhofer-entwickelt-Echtzeit-Warnsystem-11397791.html

August 5, 2026

Public NOW archive for August 5, 2026.

Sandbox with Podman and IO

Control and monitor all of your application's network traffic.

Anthropic’s New AI Model Can Identify More Software Bugs Than Ever. Microsoft Is Struggling To Fix Them Fast Enough.

This story was originally published by ProPublica. Republished under a CC BY-NC-ND 3.0 license. On an afternoon in mid-May, dozens of Microsoft engineers and their managers gathered online and in a conference room at the company’s Redmond, Washington, headquarters to discuss Project Glasswing. The tech giant was racing to fix weaknesses in its code that...

August 4, 2026

Public NOW archive for August 4, 2026.

#ki #opensource

https://the-decoder.de/machtkampf-im-silicon-valley-stoppt-vorerst-us-verbote-fuer-chinas-ki-modelle/

August 3, 2026

Public NOW archive for August 3, 2026.

August 2, 2026

Public NOW archive for August 2, 2026.

August 1, 2026

Public NOW archive for August 1, 2026.

July 31, 2026

Public NOW archive for July 31, 2026.

July 30, 2026

Public NOW archive for July 30, 2026.

#ki #hack

https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals https://www.spiegel.de/netzwelt/auch-ki-von-anthropic-griff-echte-firmen-an-a-f35a127c-dc89-4c67-afeb-b6863387cb91

#ki #forschung #sicherheit

https://www.far.ai/blog/ai-security-leaderboard

July 29, 2026

Public NOW archive for July 29, 2026.

#ki #regulierung #bundesnetzagentur

https://www.heise.de/news/Neue-Befugnisse-Bundesnetzagentur-uebernimmt-KI-Aufsicht-in-Deutschland-11383935.html

#ki #security #openAI

https://www.golem.de/news/neue-details-zum-ki-angriff-openai-hat-mehr-als-nur-hugging-face-gehackt-2607-211385.html

エージェントの登録手順をMarkdownで公開するauth.mdが良さげかも

AIエージェントがユーザーに代わってサービスへ登録するためのオープンプロトコルauth.mdの紹介です。

website-comm-template unbreaks install and closes a header-injection hole

Fixed a broken pnpm install and a no-op Prettier config, then shipped real SEO routes and fixed an email header-injection and unsafe social-link bug in the contact form.

pds-status-tophhie fixes its install and types its remote responses

Unbroke a pnpm install, replaced the untouched scaffold README with a real one, and typed the remote PDS responses instead of trusting them as any.

atproto-shortlink closes an open-redirect and header-injection hole

Redirect targets sourced from remote blue.linkat.board records were never validated. They're now checked before use, alongside a broken install, a favicon that never resolved, and a run of accessibility fixes.

linkat-directory sanitizes untrusted AT Protocol data before rendering it

Board-card URLs, avatar/banner URLs, and PDS-resolver endpoints were previously rendered or used with no validation at all. All three are now checked before use, alongside a broken-metadata fix and a broken install.

Want a police unit in your community? Build it yourselves, Mukhito tells MPs

Building police units is not the government's responsibility, Malawi Homeland Security Minister Peter Mukhito has reiterated, saying communities and Members of Parliament should take the lead in constructing the facilities while government provides technical support and police officers. With constituencies now controlling K5 billion each through the Constituency Development Fund (CDF), Mukhito inference is that...

--end-of-options

The git flag I assumed was an LLM hallucination

NOW

My current understanding.

A safer AI model still needs locks around it

This week’s releases separated prompt resistance, credential access, authorization, isolation, review, and recovery into different safety jobs.

Plumbing Homebrew into the vulnerability ecosystem

One command, six repos, three standards bodies, an advisory database, and a version comparator written in the wrong language.

OpenAI is training an attacker. 1Password is hiding the keys.

Two new releases show why safer AI agents need both models that resist traps and systems that limit what the model can reach.

When a coding agent treats silence as permission

OpenAI warned that GPT-5.6 Sol can act beyond user intent. New deletion reports show why permission has to live outside the model.

Google is putting AI agents in separate virtual machines

Google's CAPSEM puts each coding agent in an isolated VM and keeps credentials outside it. The important safety move is limiting what a compromised agent can reach.

A Route Guide for Sidecar

Implementing the gRPC Route Guide example with Sidecar.

Content addressing in package managers

Names are for humans, hashes are for everything else

Am I a target?

I keep getting traffic from the same server. It's a Hetzner Online server (supposedly) which has geodata linking it to the UAE. I've blocked the traffic, but I'm getting hella suspicious. Anyway I've been live monitoring all my traffic and I'm blocking anything that I think is even the tiniest bit sus.I'm not really thrilled...

The Helpful Bypass

Most security thinking assumes an adversary. A threat model starts with: who's trying to break in?

Unbundling the standard library

Batteries no longer included, available separately on aisle four

ターミナルの危険なコマンドを実行前に止めるTirithが良さげかも

危険な処理を実行前に防ぐターミナル向けセキュリティゲートTirithの紹介です。

The Dark Surface: Why Read-Surface Governance Can't Be Built

Every governance tool we build for AI agents—labelers, moderation systems, legal protocols, content policies—clusters on the same surface: output.

Incident Report: CVE-2026-LGTM

A series of unfortunate agents.

Google Cloud Summit ANZ 2026

Scrutineer: scanning open source without flooding maintainers

Finding the vulnerabilities is the easy part

Pattern Gates: Why Trust Architectures Break When AI Shows Up

Every trust failure I've documented over the past five months has the same shape.

Joint Guidance on Vulnerability Naming and Disclosure

Every named CVE now ships with a single-page site at .vuln.

Self-hosting email the hard way from your own routable IPv4 block up

How we refreshed self-hosted Recoil email with our own RIPE-allocated IPv4 block, and deployed Postfix/rspamd/Dovecot to get full SPF/DKIM/DMARC deliverability.

Install-script allowlists

A survey of install-script allowlist mechanisms across package managers and language ecosystems.

gittuf - a signed log for git refs

Branch protection is a row in someone else's database

Skills Registry Threat Models

How long until we see a CVE filed against a markdown file?