ls ./archive --page 12
Three crises are being bundled into one story about a company cracking. The suit against OpenAI is real and names a Chief Hardware Officer. The 630GB of iPhone 18 Pro files came out of Tata Electronics. And Siri is not still broken — Apple bought a way out for about $1bn a year.
Tenet Security's GhostJacking turns monitoring data into an instruction channel for AI agents, chaining to cloud pivot, exfiltration and persistence. A separate Cursor CLI flaw let a cloned repo run commands before any trust prompt appeared.
Shai-Hulud's second wave hit somewhere between 600 and 800 packages depending on who counted. The packages were pulled within days. The credentials it published to public GitHub repos are what produced breaches months later — including Suno's.
The claim is everywhere and the argument behind it is good. The number is not a measurement. It comes from a single blog post in May 2026 describing a local experiment on React, and the industry statistics stacked around it come from different vendors counting different things.
The 'AI slop' story says quality collapsed. HackerOne's own figures say the confirmed-exploitable rate held at about 25% through a record surge, and the share of critical and high findings went up. The problem is not slop. It is that finding got cheap and fixing did not.
Two law firms, an equine hospital, a mental health charity and three metal companies, published in batches with a two-day gap between attack and leak. It shares a name with Microsoft's Storm-1175, which is a different actor entirely.
The protest slogan is that company ethics cannot stop the AI race, only law can. The Pentagon episode is the cleanest test of that claim anyone has run — and it did not end well for the company that held the line.
For two years OpenAI told the court it could not search its own training corpus. A court-ordered deposition in April produced a database of 78 million conversations and a tool built to log regurgitation. The sanctions motion asks the court to decide the fact question without the evidence.
CVE-2026-45659 was patched in May, flagged as exploited on 1 July with a three-day federal deadline, and marked as used in ransomware on 11 August. Of 8,500 SharePoint servers exposed online, 200-plus remain vulnerable.
The viral version says a hacker exposed 2 million stolen songs. The number is one line item from one of ten sources, it is ten times smaller than what Suno told a court in 2024, and the labels had the real figure in discovery two months earlier. Underneath it is a breach Suno described as involving no sensitive personal information.
Google published Q1 2026 figures for its notification abuse work: automatic permission revocation for inactive sites, HTTP 429s past a threshold, and network-level analysis of coordinated service workers.
The annotation feature opens a direct channel between viewers and the person sharing. A Security's 'Zoomsday' chain starts with a missing bound check in the text annotator and ends in remote code execution.
The Water Cyber Shield Act would give the EPA explicit cyber authority and $300m a year. Separately, DEF CON Franklin and the rural water association launched a monitoring centre funded by Craig Newmark.
Sequoia, Khosla and Coatue backed a 2025-founded company whose pitch rests on one test: general-purpose models could run end-to-end attacks and failed to defend against the same attacks.
'AI Sidebar with DeepSeek, ChatGPT, Claude and more' had 300,000 installs and a 4.6-star rating before Google removed it in January. It returned in August, clean for eleven days, then shipped a 21-line addition.
Flight 591 out of Las Vegas carried 199 passengers and a network called 'Delta WiFi Fast' that served a page harvesting Google logins. Crew killed the Wi-Fi for 30 minutes; federal officers boarded on landing.
QUIRSO traced exploitation of CVE-2026-59310 across 47 countries, with victims first contacting attacker infrastructure on 3 August. The persistence mechanism is a cron job running an open-source reverse SSH tool.
Unit 42's analysis of Kimwolf v7 describes HTTP/2 floods that mirror real browser behaviour at the protocol and header level. It runs on Android TV boxes with port 5555 open, and resolves its C2 through Ethereum Name Service.
CERT-UA describes UAC-0145 posing as recruiters for a real consulting firm, running Zoom interviews, then handing over a WireGuard config that fails on purpose — so the candidate downloads 'SopraVPN' instead.
Microsoft's analysis describes two Polygon wallet addresses holding proxy URLs and leak-blog posts, with negotiation over Session. There is no hosting provider to serve, because there is no hosting.
Three outlets counted this Patch Tuesday three different ways, which tells you how little the total matters. CVE-2026-68820 in afd.sys is a use-after-free rated 7.0, Check Point ties it to Lazarus, and the chain starts with a fake job offer.
Two poisoned releases in March, published with a token stolen in the Trivy breach and kept alive by an incomplete rotation. CloudSEK's dataset of 434,000 captured files now maps possible exposure across 2,500+ organisations.
GPT-5.6-Cyber completes 95.0% of requests on OpenAI's advanced cybersecurity evaluation, against 1.5% for the base model. It is not on the API. You apply to Daybreak Red, and someone decides.
Genians found the stack on infrastructure linked to the North Korean espionage unit. Running models offline means stolen documents never touch a provider, and no usage policy applies. That is the point.