Search the archive

challenges of AI adoption. Claim your free copy ›

OPENSCREEN

OpenScreen is a free, open-source desktop recorder for Windows, macOS, and Linux that captures screens without watermarks or subscriptions. It uses GPU-accelerated recording and is designed for quickly producing product demos and shareable videos.

CVE MCP SERVER

CVE MCP Server gives Claude 27 security-intelligence tools backed by 21 APIs, including CVE lookup, EPSS scoring, CISA KEV, MITRE ATT&CK, Shodan, and VirusTotal. It packages vulnerability research and threat-enrichment workflows behind MCP so agents can gather and correlate security data from one server.

CLAUDE ADS

Claude Ads is a Claude Code skill for paid-media operations across 12 advertising platforms, including Google, Meta, YouTube, LinkedIn, TikTok, and X. It supports source-grounded audits, deterministic scoring, versioned JSON reports, and capability-gated account changes.

HAVE THE FRONTIER LABS MIXED UP AI SAFETY AND SECURITY?

Frontier labs may be treating security like probabilistic AI safety, accepting low failure rates where deterministic containment should be required. Recent agent escapes show that weak sandbox rules and ignored alerts can defeat otherwise useful monitoring, so known controls must work every time and trigger real intervention.

7 AI MODELS RAN REAL BUSINESSES: $12,431 IN FAKE INVOICES, 2,797 SPAM EMAILS, $0 REVENUE

Seven frontier models received $300, an unlocked Mac mini, business tools, and 72 hours to make money. They generated no revenue while spending nearly $3,200, sending 2,797 emails, and issuing $12,431 in unsolicited invoices, exposing risky behavior and weak business judgment.

USING GPT-6 ASTRA INSIDE CLAUDE CODE IS THE NEW META

A free local model-gateway plugin routes GPT requests from Claude Code through a ChatGPT login while leaving Claude requests on a claude.ai account. The workflow uses Astra to orchestrate, cheaper GPT-5.6 models to execute, and Opus to review, with setup notes for the context-window suffix, finite allowances, and proxy breakage.

related to payment authorization and UI. Watch now.

YOUR AI CAN FIX THE BUG. BUT CAN IT FIND IT?

Experiments on hidden Django bugs show cheaper models can follow detailed instructions and reproduce known failures, but struggle to invent the questions and adversarial inputs required to discover defects independently.

WHAT WE CAN LEARN FROM CLAUDE'S FABLE 5.1 SYSTEM PROMPT

A comparison of Fable 5.0 and 5.1 system prompts shows how changing model quirks, product policies, and tool-use guidance can create prompt debt unless instructions evolve with the model.

THE QUESTION WAS ALREADY WRITTEN

Anthropic's machine-generated Lean proof of Fermat's Last Theorem is mechanically verified, but its 13 million lines are too large for most consumer hardware and too opaque to serve as reusable mathematics.

NVIDIA PERSONAL AI ROUTER - PAIR

PAIR routes independent local-inference requests across compatible computers on the same network, exposing Ollama- and OpenAI-compatible endpoints while keeping prompts and responses local when every configured component is local.

response and limit business impacts. See how →

N-ABLE PATCHES MAX-SEVERITY N-CENTRAL FLAW AMID ONGOING ATTACKS

Unauthenticated attackers can execute code on internet-exposed N-central RMM servers through a low-complexity attack exploiting CVE-2026-86218. Huntress has flagged it as a potential zero-day after finding a compromised customer instance whose logs had already rotated. Shadowserver counts nearly 1,500 exposed N-central servers, most in the United States and Europe. On-premises operators need N-central 2026.3 Hotfix 4 immediately, since HF3 systems remain vulnerable to this flaw despite last week's fixes for the authentication-bypass pair CVE-2026-86206 and CVE-2026-86207.

POSTGRESHELL: THE DATABASE POWERING MUCH OF THE INTERNET HAD AN OPEN DOOR FOR 12 YEARS

CVE-2026-6471 affects PostgreSQL 9.4 through 18. A low-privilege REPLICATION account could supply an unchecked output-plugin path, load attacker code, gain PostgreSQL superuser access, and establish persistence, while Windows targets could fetch a DLL over SMB. It is recommended to patch PostgreSQL, remove unused replication privileges, restrict pg_hba.conf, block outbound SMB and NFS, and monitor suspicious replication slots.

MATHSPACE DISCLOSES DATA BREACH AFFECTING OVER 1M PEOPLE

Online math learning software Mathspace disclosed a data breach affecting over 1M students, guardians, school staff, and Mathspace staff from Australia and New Zealand. The attackers exploited a vulnerability in a self-hosted Metabase instance used for internal reporting. Mathspace did not disclose exactly what data was breached but stated that academic records, authentication credentials, or any data linking records to a school (other than potentially email domains) were not leaked.

FILELESS ELF EXECUTION VIA O_TMPFILE

A red-team loader chains O_TMPFILE with execveat(AT_EMPTY_PATH) to run an ELF that never gets a directory entry, allocating a real inode on /tmp with no dentry so ls, find, inotify, and fanotify never see it, then reopening the fd read-only through /proc/self/fd/ to clear ETXTBSY and executing it directly with no path string ever constructed. The technique sidesteps fileless detection built for memfd_create because it produces none of the expected artifacts: no memfd: prefix, no 00:01 device, no syscall 319, and the only residual signal is the (deleted) suffix that countless legitimate open-unlink-exec programs also generate. Defenders should stop keying fileless detection on memfd_create specifics and instead watch execveat with AT_EMPTY_PATH against anonymous /tmp/# (deleted) executables, correlating prctl(PR_SET_NAME) renames and loaders that unlink /proc/self/exe, since this combination is absent from Elastic's FENIX 15-technique matrix.

TAMPER-EVIDENT AUDIT LOGS: HASH CHAINS, MERKLE TREES, AND EXTERNAL ANCHORING

This article explains how access control addresses the key questions an auditor has about logs by introducing specific mechanisms: hash chains for integrity, Merkle trees with signed checkpoints for an append-only history, and trusted time anchors like RFC 3161 or OpenTimestamps. It provides an implementation perspective on creating a cryptographically verifiable audit trail, including data structures, proof algorithms, and working Rust code. The guidance is based on established standards.

WIZ RED AGENT FINDS ITS WAY INTO SNOWFLAKE'S INTERNAL JIRA THROUGH A FLAW IN A GITHUB COPILOT-ASSISTED PR

Wiz Research's Red Agent autonomous security research tool discovered a vulnerability in a GitHub Actions workflow in the snowflakedb/snowflake-connector-net repository that allowed attackers to gain access to Snowflake's internal Jira. The vulnerability stemmed from a regression introduced in a PR that was reviewed and approved by GitHub Copilot, enabling code execution via template injection. Users of AI-assisted code generation tools should rigorously review code for introduced vulnerabilities. In this case, Copilot introduced a vulnerability by replacing a safe env + jq parsing pattern with direct string interpolation.

measurable, controlled outcomes. Register for Thursday's event.

RAPTOR

RAPTOR is an autonomous security research framework built on Claude Code that integrates static analysis, binary analysis, LLM-driven vulnerability validation, exploit generation, and patch writing into a single workflow. It ships over 200 custom Semgrep, Coccinelle, and CodeQL rules, a six-stage exploitability validation pipeline that filters scanner noise before a finding is called real, and optional Z3 pre-screening that drops provably unreachable dataflow paths. Subprocesses that handle untrusted repositories are sandboxed using Linux namespaces, Landlock, and seccomp, and the analysis layer accepts Anthropic, OpenAI, Gemini, Mistral, or local Ollama models with a per-run cost cap.

DAYBREAK FOR FRONTLINE DEFENDERS: $1B TO PROTECT ESSENTIAL SERVICES

OpenAI has committed $1 billion in subsidized Daybreak access, training, and technical support for resource-constrained defenders, with consumption targeted over the next six months and prioritizing water and wastewater utilities, electric grid operators, state and local governments, community banks, nonprofits, and open-source maintainers. Daybreak Blue covers routine defensive work on mainline models. Daybreak Red provides approved organizations with specialized cyber models. Roughly 2,000 organizations have already onboarded, and more than 35 partner products are shipping through the Daybreak Defense Network.

BEYOND LAZARUS: HOW NORTH KOREA ORGANIZES ITS CYBER OPERATIONS

North Korea has been shown to run offensive cyber as a self-funding arm of the state, with theft revenue estimated at over a billion dollars annually feeding sanctioned weapons programs. Sekoia and Kudelski Security mapped the hierarchy behind that model, covering the GRIB (ex-RGB), the NIA (ex-MSS), renamed in June, and the decomposition of the Lazarus umbrella into APT38, Andariel, Famous Chollima, and the Sleet sub-clusters. The paper also traces the enabling layer of IT worker units, Chinese and Russian academic pipelines, and front companies such as Ryonbong and Chinyong, while cautioning that constant institutional reshuffling limits how long any DPRK org chart stays accurate.

JSCEAL MALWARE CAN BYPASS GOOGLE AUTHENTICATION USING STOLEN SESSION COOKIES

JSCeal is distributed via fake TradingView installers advertised on Facebook and Google. It harvests browser passwords, cookies, OAuth tokens, and profile information from Chromium browsers. Stolen Google cookies can restore authenticated sessions and bypass login screens. The malware also logs keystrokes, takes screenshots, and sets up a local proxy that manipulates traffic for Binance, Bybit, Ledger, and various other services.

HACKERS DRAIN $320M IN BITCOIN FROM LIQUID NETWORK, CLAIM THEY'RE THE GOOD GUYS

Attackers withdrew about 4,000 BTC, roughly 95% of Liquid's federation wallet. Liquid disabled bridge nodes and asked exchanges to halt L-BTC deposits and withdrawals. The attackers used SideSwap's peg-out process, though Liquid says no Peg-out Authorization Key was compromised. They claim they will return most funds after every node receives a fix.

practical AI, security, and networking guidance. Register Now

CONNECTWISE WARNS OF NEW SCREENCONNECT FLAW WITHOUT PATCH

ConnectWise disclosed an unpatched file transfer flaw affecting both cloud and on-premises ScreenConnect deployments.

MANCHESTER AIRPORTS GROUP DATA ON 8.8 MILLION PEOPLE LEAKED AFTER RANSOM REFUSAL

FulcrumSec leaked about 550GB allegedly stolen from Manchester Airports Group after MAG declined a ransom.

STATEMENT: NOTICE OF PERSONAL DATA BREACH AFFECTING FORMER AND CURRENT EMPLOYEES

Natural Resources Wales disclosed a spreadsheet containing diversity-monitoring data for people employed between April 2013 and March 2018.

Get 50% off the #1 rated password manager today

CLAUDE FABLE 5.1 ON AWS COMES WITH NEW DATA CONTROLS

Claude Fable 5.1 is available through Amazon Bedrock and Claude Platform on AWS, with prompts and outputs subject to retention for up to 30 days and potential safety review by Amazon personnel. Planned Enterprise Frontier Safeguards will let eligible customers keep monitoring data in cloud infrastructure they control.

SOPHOS BRINGS OPENAI MODELS TO VULNERABILITY PRIORITIZATION

Sophos is developing Exploit Path Verification to assess whether attackers can actually exploit vulnerabilities given an organization's network access, privileges, patches, and security controls. The capability will use OpenAI cyber models with analyst review, but availability has not been announced.

HAVE THE FRONTIER LABS MIXED UP AI SAFETY AND SECURITY?

Frontier AI labs confuse probabilistic AI safety with deterministic security, treating prompt-injection defenses as largely solved despite high failure rates. False positives, weak containment like proxy and firewall gaps, and reliance on detection over containment enabled the failures.

A DECADE OF SOFTWARE INFRASTRUCTURE

The next decade will be defined by software infrastructure as AI turns everything into infrastructure. This post outlines six megatrends, including AI-rebuilt development lifecycles, infrastructure re-architected for agents, and APIs replacing apps. Agent-driven buying and early-stage machine-learning infrastructure are predicted to expand opportunities for infrastructure startups.

RUN CLOUD AGENTS ON MACHINES YOU MANAGE

Cursor now lets cloud agents run on your own infrastructure, while Cursor still handles the agent loop, inference, and orchestration. Self-hosted workers can scale dynamically across AWS, Vercel, Modal, Cloudflare, Macs, etc., which gives teams direct access to internal systems, custom hardware, and existing build environments.

OPENSHIFT ADDS A UNIFIED SECRETS MANAGEMENT CONSOLE

Red Hat's new console plug-in brings certificate and external secrets resources into one interface, helping administrators investigate expiration warnings and failed dependencies. It is available as an administrator-only technology preview for OpenShift 4.22.

EXPLAINING 3PL

A third-party logistics provider holds a brand's inventory and handles picking, packing, shipping, and returns while the brand keeps product and customer ownership. This explainer details its rate-card pricing, EDI and API integrations, and cut-off times, arguing that warehouse location and performance directly set delivery dates in Shopping ads and marketplace badges like Prime. It also notes rising fee and mandate disputes, including Amazon surcharges and conflicting accounts of TikTok Shop's shipping mandate.

FINOS INTRODUCES PORTABLE RECORDS FOR AI-ASSISTED WORK

The konspekt project proposes an open format for preserving project decisions, artifacts, and source provenance across AI tools. Records live in plain files and require human acceptance, creating an audit trail independent of any model provider or chat platform.

DOES CONCENTRATION RISK MATTER IN AI?

Heavy revenue concentration among AI labs and across the AI stack does not necessarily give large customers leverage - vulnerability is accessed by layer.

AMAZON LINUX 2027 ENTERS PUBLIC PREVIEW

The preview includes kernel 7.1+, SELinux enforcement by default, updated development tools, and AWS Neuron driver support.

AI SPENDING OUTPACES PROOF

Enterprises are experimenting with budgets and model routing while struggling to measure returns.

THE BIG WINNERS AND LOSERS IN US WEBSITE TRAFFIC OVER THE PAST YEAR

Google and YouTube still dominate web traffic in the US, but AI is reshaping the landscape. ChatGPT traffic jumped 48.4% year over year to 1.09B monthly visits, while Bing fell 50.4%. Meanwhile, GA4 is losing visibility into AI-driven traffic, with “direct” traffic increasingly absorbing visits it can't attribute.

STREAMING IS RAISING PRICES FASTER THAN CABLE EVER DID

Streaming prices have risen 3X faster than inflation since 2022. The 8 major services now cost around $151/month without ads or bundles, up from $90 four years ago. As subscription fatigue grows, Gen Z's “streaming cycling” trend is gaining traction. Consumers subscribe to one service at a time, then cancel and switch to another.

THE BEST ADS DON'T CREATE DESIRE, THEY BORROW IT

The best copy doesn't create desire. It taps into wants customers already have, then connects the product to those deeper motivations. Identify the emotional driver behind the purchase, name the obstacle standing in the way, and show the customer what success looks like through a vivid, specific scenario. Rather than simply stating a benefit, make readers experience it before they buy.

SHE PAID PEOPLE TO CANCEL HER COMPETITOR

Cursor turned a giveaway into a competitor-poaching tool by offering its $200/month plan free to customers who canceled a rival and proved it with a screenshot. Claiming meant DMing the employee running the offer from her personal account, creating a private channel for retention offers and switching feedback. A second run added a 24-hour deadline and converted even better. Customers also posted cancellation screenshots publicly, creating free marketing.