In wger, a trainer with no gym could read notes of other no-gym users, advisory says

Hammond: criminals can skip scam brands, as uncensored models are a free download

Warden infostealer reportedly targets Claude Code and AI coding tool keys

Latest intelligence

Google's EmbeddingGemma 2 puts video and audio search on the device

Google says the full model needs about 567MB of RAM when quantized on a Pixel 11 Pro. Its index is data to govern.

The AI-First Web

TII's test: rival AI models often reply in formal Arabic to Emirati questions

Security & Trust

ASOS says attackers misused its customer messaging platforms

The AI-First Web

One voice-AI builder says buyers want the power of a human, the control of a robot

Innovation & Growth

AWS shows how AI data agents can query as the real user, not a shared role

Innovation & Growth

More cables do not mean more backup when all run through one strait

Security & Trust

Building digital trust starts with infrastructure. The security posture of your framework is the foundation.

All stories →

In wger, a trainer with no gym could read notes of other no-gym users, advisory says

An advisory describes an access check that treats two empty values as a match, for unassigned trainers and unassigned users

Hammond: criminals can skip scam brands, as uncensored models are a free download

John Hammond showed criminal AI brands that looked like scams, then a forum guide to running uncensored models anyone can download.

Warden infostealer reportedly targets Claude Code and AI coding tool keys

Version 1.9 reportedly targets Claude Code and Codex CLI files, so an agent's login is only as safe as the machine.

10M+

sites analysed

888

insights published

frameworks tracked

regional editions

The AI-First Web

The web is being rebuilt for machine consumption. What that means for every business building online.

All stories →

$OpenAI posts 372 AI math results; the hard part is checking them$

OpenAI posts 372 AI math results; the hard part is checking them

Cheap machine output moves the bottleneck to review. Mathematics is an early test of that problem.

Common Sense Media: ChatGPT parent alerts missed crisis chats on new accounts

Testers on new parent-linked accounts got no alerts. OpenAI disputes the method. The trigger is the real question.

One voice-AI builder says buyers want the power of a human, the control of a robot

Text replies and citations give firms control; the transcript comes last so replies stay fast.

Cost of Legacy

The hidden costs holding businesses back. How modern infrastructure unlocks speed, savings, and scale.

All stories →

AssemblyAI could not fix its bought support bot fast enough; it solved about 10%

Matt Lawler says a bought bot solved about 10%; AssemblyAI's own agent reached 80% in week one, by its account.

Microsoft and Google back a shared format for business metric definitions

Apache Ossie aims to let companies define 'revenue' once and reuse it across analytics and AI tools.

UK Civil Service Moves Cyber Governance From Mandates to Services

Civil Service Deputy CISO: a 2025 audit found "no proper implementation plan" three years into the 2022 strategy.

Modern Stack

The technologies enabling the next wave of digital business. What leaders are building on and why.

All stories →

Checking AI output, not making it, is now the bottleneck, a voice AI show argued

A speaker on a voice AI show argues the hard part is auditing agent output, and the show's guest works for a vendor.

AWS shows how AI data agents can query as the real user, not a shared role

A new AWS pattern keeps the user's identity token away from the AI model and lets existing data rules apply.

More cables do not mean more backup when all run through one strait

Google reportedly paid two to three times the going rate for a route that avoids the Red Sea. Other cloud giants are reportedly looking too.

Framework Migration

Moving from legacy to modern isn't just a tech upgrade. It's a competitive advantage.

All stories →

Kubernetes v1.35 stops nodes on cgroup v1 from starting by default

Security & Trust

Wiz researcher warns coding agents can add dependencies developers may not know about

AWS design moves machine identity keys and records onto managed services

Atlassian says AMP tags AI code; IDC analyst says proof of review matters more

Anthropic widens vetted access to its cyber AI as flaw counts pass 100,000

Airtel moves enterprise messaging fraud checks into the phone network

When exploits come fast, exposing fewer things is the first defence

Security & Trust

Victorian student data breach traced to one school's unpatched server

A suspected ransomware attack halted classes at a Japanese university

Microsoft's on-premises patch release hit a record near 1,000 flaws in September

Business Efficiency

AssemblyAI could not fix its bought support bot fast enough; it solved about 10%

Microsoft and Google back a shared format for business metric definitions

UK Civil Service Moves Cyber Governance From Mandates to Services

Innovation & Growth

DNS root key is set to change October 11; resolvers must trust it first

GitHub says agents are pushing code so fast it is rebuilding Git storage

Cloudflare's new beta traces supported steps a request takes inside its network

The AI-First Web

One Copilot CLI model sent out secrets in 50% of Adversa's test runs

Open-source gateway keeps AI agent credentials out of config files

VB Pulse: 67% run, pilot or build a semantic layer; 13% call it primary source

Security & Trust

AWS offers Z.ai's GLM 5.3 to enterprises, with a security-testing agent as demo

Red Hat says it fixed 400+ novel Java library flaws and sells old-version fixes

Datadog finds stolen-AWS-key tools test for Bedrock AI access

Mistral says its new open model does security work rivals refuse

GitHub Enterprise flaw turned a push option into server code execution

Exploited NetScaler flaw can lock out a company that uses SAML sign-in

A clean transcript can hide a voice call that went wrong

A passing AI test proves little until the judge is checked against people

arXiv caps submissions as AI makes bad papers cheap to write

AI agents are being blocked, and users cannot tell if the site meant it

On debated questions, AI answers can change with who the model thinks is asking

Kubernetes node swap fit up to 3× more AI sandboxes per node in tests

Survey: AI agent security sits with model vendors; many agents share logins

At AssemblyAI, once an AI agent answers first, the human job becomes judging it

Appeals court: the user, not Perplexity's AI agent, 'accessed' Amazon

Manus agents get their own wallet and phone number in a personal app

Airbnb's CEO says chatbots fit travel poorly and agents lack a platform

Senate hears how OpenAI's test agents breached Hugging Face

As AI Scam Defense Reaches Consumers, Legacy Web Infrastructure Lags Behind

WordPress 7.0 Ships — Then Immediately Starts Migrating Its Own Admin to React 19

NIS2 Audit Deadline: June 30. Every Unpatched WordPress Plugin Is a Compliance Violation Worth 10 Million Euros.

Drupal Core SQL Injection: Anonymous Access, CISA KEV, Exploited in the Wild