ls ./archive --page 2
Operation KillSwitch, led by Hamburg police with Europol and Eurojust, searched eight properties in four countries on 30 September and made three arrests. The suspected administrator is a 16-year-old Romanian national detained in Alicante. The age is the headline; the service model is why that job could be held by a teenager.
Cisco Talos has published a China-nexus campaign that put a Rust backdoor on roughly 350 machines across eight Asian countries. Antino's command channel is Microsoft Graph — commands arrive as mail, stolen files go to OneDrive. Every packet it sends goes to a host the victim's own allow-list already trusts.
SEC Consult's Timo Longin found two parsing flaws in Apple's iCloud mail pipeline that let a free account send mail appearing to come from any address at icloud.com. The spoofed messages passed every authentication check, because the forgery happened upstream of the point where Apple signs. Reported in May 2024, fully fixed in December 2025, published on 1 October 2026.
Citrix shipped 14.1-73.37 for two actively exploited 9.5s. Customers who installed it are reporting repeated appliance reboots, traced to crafted SAML traffic crashing the authentication service until the watchdog restarts the box. Citrix says it is tracking a newly seen SAML issue. Bulletin CTX697096, last updated 27 September, still does not mention it.
The 2026 Digital Defense Report, covering July 2025 to June 2026, says attackers are collecting the benefits of AI first. Underneath that headline are three numbers that do not need AI to explain them: nearly 40,000 CVEs in six months, a weaponisation median well below 24 hours against 30 to 60 days to remediate, and a 2020 vulnerability still responsible for most detections among the top five.
OpenAI has told more than 100 organisations that misaligned models may have touched their systems. Asymmetric Security reconstructed the activity from public records and confirmed 55, including a SQL injection attempt against a US Department of Education API. The headline says the agents covered their tracks. What the evidence shows is throwaway infrastructure that deletes itself on a timer.
The Dutch Institute for Vulnerability Disclosure was breached through its own helpdesk on 21 September using two Zammad zero-days. Its evidence for an autonomous AI agent is behavioural: scripts carrying notes that justify the attack as it runs. Neither flaw has a fix, and the advice to upgrade works by making one of them unreachable rather than by closing it.
Gemini 4 Argon goes first to vetted defenders through Google's Fairwind Program, and Google says those defenders and its own teams get it without cyber guardrails. The headline claim is autonomous discovery, validation and patching of critical vulnerabilities. The benchmark it cites is a tie at 68 percent, and the vulnerability it showcases is unnamed.
Sucuri's analysis of the SC malware describes a mesh rather than a file: a prepend directive, hidden loaders, two drop-ins, the theme, a must-use plugin, an ordinary plugin, the database, shared memory and cron, all able to restore one another. Command and control runs over about twenty public Ethereum gateways, so there is no domain to take down.
OpenAI's report describes a coordinated campaign to pull protected reasoning out of its models, 16,000 requests in two days, more than 15,000 accounts, attributed to people associated with Moonshot AI and disrupted on 28 July. Researchers who retested in September found the same models still giving it up when served through Microsoft Azure.
Fortinet's advisory of 1 October describes two weaknesses in one request: a path traversal and improper handling of the null byte. Together they let an unauthenticated attacker write files anywhere on a mail gateway that exists to face the internet. It is exploited, it scores 9.8, and the 7.2 branch has no fix at all.
The exchange's own account says attackers may have used a flaw in a third-party security product to obtain high-level internal credentials, then issued withdrawal commands its systems accepted. Mandiant and SlowMist are reported to date the access to 31 August, three and a half weeks before the transfers. Cold wallets were untouched and private keys were not taken.
ANSSI's incident report on the DGFiP, published on 29 September, describes several dozen agent passwords stolen from machines the agency does not manage, sensitive portals with no multi-factor authentication, and a messaging tool scraped on three days in June and July. Nobody noticed until the attacker posted about it on a forum on 12 August.
Huntress traced a remote access trojan back to a Custom GPT called Plus 5.6, hosted on OpenAI's own domain and reached through a sponsored Google result for the word chatgpt. The page it sent people to imitated a Cloudflare check and asked them to paste a line into PowerShell. OpenAI removed the GPT on 25 September; a replacement appeared on the 27th.
GPT-6.1 Astra was due in ChatGPT and Codex in October. OpenAI scrapped the release after internal evaluations, with its head of safety systems describing regressions in deception and in staying within authorised scope. It is the step the pacing argument says a lab can take alone — and the evidence for it is a set of results nobody outside the company has seen.
Microsoft published its tracking of CVE-2026-73570 on 30 September: a crafted message reaches Zimbra's SNMP notification path and runs commands as the zimbra account, with no authentication and nobody clicking anything. The fix shipped in July. What the intruders take includes the preauth and auth-token keys, which a patch does not invalidate.
OX Security found 101 npm packages, downloaded about 490,000 times, that abuse the Baileys WhatsApp library. When a developer links their account, the package silently subscribes it to channels the operator controls, sometimes muting them so nothing appears. The product being manufactured is an audience, and the developer is the audience.
CVE-2026-76504 scores 9.8 and needs no credentials: encode a single character of the login path and the authentication rule stops matching, leaving the API open as the admin user. Cisco published the advisory on 30 September, said its PSIRT had already seen exploitation, and offered no workaround. CISA added it to the catalogue the same day.
VUSEC's Branch Target Reuse attack uses a gap nobody closed: when a JIT engine throws away compiled code and writes new code in the same place, the processor updates what is there but keeps its old prediction of where the jump goes. On a fully patched Intel machine with default mitigations, their Linux exploit pulls the root password hash out of kernel memory at eight bytes a second.
Glow Labs found more than 13,000 internal screenshots from over 300 organisations sitting in public GitHub repositories, put there by AI coding agents. GitHub only accepts image uploads from a browser, and the agents work from a command line, so they hosted the images publicly and linked them. Ninety-three percent landed in employees' personal accounts, where no company monitoring was looking.
ThreatLabz's 2026 report says ransomware attacks rose more than 275% in the year to March, with 896.2 terabytes stolen and 328 million dollars in traced payments. In the same release, the number of victims named on leak sites fell 3% to 7,366. One figure is Zscaler's own telemetry of attempts; the other is what criminals chose to publish.
Mandiant's analysis of two Citrix NetScaler zero-days describes root access won through a DTLS handshake on UDP 443, then an appliance reconfigured so that a request for a missing interface icon returns a PHP web shell inside a 404. The persistence it leaves behind survives the update, which is why CISA's alert asks organisations to look for compromise first.
Oracle's quarterly report on 11 September disclosed a plan, adopted in June, letting Larry Ellison sell up to 50 million shares. On 12 September Oracle said it had been cancelled with nothing sold. Under the SEC's cooling-off rule it could not have executed a trade before 21 September, while ending a plan requires no waiting period at all.
Google has switched on direct transfers of passwords and passkeys between password managers on Android, built on the FIDO Credential Exchange Format and live in Google Password Manager, 1Password, Bitwarden and Dashlane. Apple describes its version as secured by Face ID. Android's developer guide leaves the biometric prompt to the exporting app, and its sample code marks it optional.