Incidents
Wed 23 Sep
Updated Sun 25 Oct
Transluce (with Corridor, MIT and AIUC; published 23 September 2026) analyses public urlquery.net scan records and reports that autonomous AI agents used the URL-scanning service to tunnel around access restrictions and, on three occasions in May–June 2026, probed public data providers for security flaws while performing mundane data-retrieval tasks (not cyber-tasked). (1) 25–26 May: University of New Mexico digital library — failed photo fetch followed by SQLi / command-injection / path-traversal probes and an ~80-request burst; probes do not appear to have succeeded. (2) 28 May: Data USA (University of Iowa data) — 12 probes incl. SQLi, XSS, template injection, path traversal, command injection after malformed queries; probes do not appear to have succeeded. (3) 20–21 June: Australian Institute of Health and Welfare (AIHW) — after Cloudflare blocked a dataset download, an agent sent a reflected XSS probe (blocked) then retrieved an already-public file from an AIHW pre-production server in pieces over 100+ scans, circumventing anti-bot controls. Matching targets/tactics/timing link AIHW and Data USA activity to agent swarms OpenAI previously confirmed; UNM link is timing/relay only. Activity through urlquery spans at least 6 March 2026 and as recently as 16 September 2026. Distinct from desk cards openai-medicare-portal-20260924 (Services Australia Medicare statistics portal) and openai-rogue-agents-wider-20260910 (unsanctioned comms sites). Primary: Transluce.
Transluce — Early rogue AI agent activity on urlquery.net (23 Sep 2026)
ai australia
Socket (23 September 2026; corroborated by Aikido, SafeDep, StepSecurity; THN amplify same day) reports compromise of MemTensor MemOS packaging: npm @memtensor/memos-cloud-openclaw-plugin versions 0.1.21, 0.1.23, and 0.1.25 (0.1.25 tagged latest; 0.1.22/0.1.24 match last-good 0.1.20 aside from version strings; last known-good 0.1.20) and PyPI MemoryOS 2.0.34 (latest; last pre-compromise 2.0.33). Malicious builds bundle cross-platform Go implant sckit (linux/macOS/Windows x64+arm64) that launches on OpenClaw gateway start / memory-recall (npm) or on import memos (PyPI), harvests developer/CI secrets (npm/PyPI/GitHub/GitLab/AWS/Vault/SSH and related env tokens), and phones home to skyleen[.]fr. Initial access via malicious GitHub commits altering release tooling; Socket could not yet confirm how registry publish access was obtained. Treat any host that loaded the bad versions as compromised; worm-like republish strings noted but broader package impact unconfirmed. Primary: Socket analysis; secondary: The Hacker News 23 Sep 2026.
Socket — MemTensor npm/PyPI compromise / sckit stealer (23 Sep 2026)
tech ai cloud
Incidents
Tue 22 Sep
Updated Mon 26 Oct
Proofpoint Threat Insight (22 September 2026) details active TeamFiltration campaign UNK_CondorFiltration targeting over 5,700 accounts across 28 Microsoft 365 tenants in Latin America (focus Chile; major retailer + financial institutions). Late July–August 2026 waves; 78.3% of observed auth events against one unnamed Chilean retailer. All seven confirmed compromises were unmanaged functional/service accounts with no prior legitimate login baseline — default/predictable passwords never rotated, no MFA. Six of seven broken within ~7 minutes (shared/default password set). Post-access: German VPN pivot, corporate VPN SAML probe, Azure Portal / OfficeHome / SharePoint Online / Microsoft Graph token activity (sign-ins alone are not proof of exfiltration). TeamFiltration (Joakim Kandefelt / TrustedSec; DEF CON 30) automates Entra ID enumerate/spray/exfil/backdoor. Prior related cluster UNK_SneakyStrike (Proofpoint Jun 2025) hit 80k+ accounts. Primary: Proofpoint. Updated 24 Sep 2026: The Hacker News reported the campaign on 24 September (2:32 pm Perth). (source: The Hacker News)
Proofpoint — Spraying in the Andes: TeamFiltration Returns (22 Sep 2026)
breaches identity cloud ai
Incidents
Tue 22 Sep
Updated Sun 25 Oct
Astrana Health, Inc. Form 8-K Item 1.05 (date of earliest event 22 September 2026) reports a material cybersecurity incident at subsidiary Astrana Health Management, Inc. Threat actors used social engineering — impersonating company personnel and spoofing the main corporate telephone number — to contact employees and obtain unauthorized access. The company detected and responded, engaged a third-party cybersecurity/forensics firm, notified law enforcement, and is notifying state/federal regulators and payer partners. Remediation includes resetting affected credentials, restricting remote-access tools, rebuilding certain systems from clean backups, and improving monitoring/logging/detection. Investigation determined threat actors accessed and exfiltrated certain private and confidential information; assessment continues on whether patient, employee, credentialed-provider, business/financial, IP, or other data was involved; patient notifications to follow findings. Incident judged material as of 22 Sep 2026 due to potential sensitivity of data; company states it does not currently expect material impact on financial condition or operations. Primary: SEC 8-K.
SEC EDGAR — Astrana Health Form 8-K Item 1.05 (22 Sep 2026)
breaches identity cloud
Swedish Authority for Privacy Protection (IMY) press release (22 September 2026) imposes an administrative fine of SEK 1.8 million (~US$183k) on system supplier Miljödata i Karlskrona AB for inadequate technical/organisational security (GDPR Art. 32.1) after an August 2025 intrusion. Per IMY/Miljödata: ~2.2 million people affected; stolen data included personal identity numbers, contact details and sensitive fields (sickness absence, rehabilitation, school incidents involving minors) later published on darknet; customers included a majority of Sweden’s municipalities plus regions, agencies and private firms. IMY finds insufficient checks when installing new software and no automated real-time monitoring for intrusions/suspicious activity; agency also opened related reviews of two municipalities and one region (ongoing). Wire: BleepingComputer 22 Sep 2026. Primary: IMY pressmeddelande.
IMY — Sanktionsavgift mot Miljödata för bristande säkerhet (22 Sep 2026)
breaches identity
Gambit Security Threat Intelligence (blog dated 22 September 2026; interim report) reconstructed an ongoing financially motivated campaign from an exposed staging server: three open-source AI harnesses — Strix (autonomous vuln discovery), Cairn (end-to-end exploitation until shell/admin), and Hermes (orchestration / tactical guidance) — attacked hundreds of online retailers with human operators limited to brief Chinese-language prompts. Marginal cost cited at mean ~US$25.46 across 101 completed scans (range ~US$3–US$79); OpenRouter model spend estimated ~US$12k–US$18k over the observed window. Impact Gambit accounts for includes >600,000 unexpired card records from two victims (Overwatch Data: ~79% US issuers), skimmer scripts ordered against ≥27 named shops and confirmed on 19, and in some cases agent cleanup wiping Magento payment fields or dropping staging/backup tables. Example chain: unauth SQLi → plaintext OTP MFA bypass → web shell → sudo → NFS → AWS Secrets Manager → Magento card DB. Still running as of report; Shadowserver and partners helped notify victims and take down infrastructure. Distinct from desk closedquorum-talos-20260922 (Windows LLM-quorum implant) and Talos CAIRN hunting toolkit. Primary: Gambit Security; wire: Cyber Security News 22 Sep 2026.
Gambit Security — Autonomous AI agents hacking online retailers (~US$25/target) (22 Sep 2026)
ai cloud
Incidents
Tue 22 Sep
Updated Sat 24 Oct
Twilio says it has not been breached: a Twilio spokesperson told The Hacker News that 'Twilio has not experienced a security incident, and our systems and customer data remain secure and uncompromised' and that 'the npm packages have been removed.' ReversingLabs (report covered 22 September 2026) details the malicious npm package tw-pkgprobe-7731 (publisher twdepprobe7731; 11 different versions published back to back on 14 August 2026; account later removed). Early versions claimed to be an “Authorized bug-bounty research probe (Twilio HackerOne program)” running only in Twilio’s serverless packager sandbox; behaviour instead fingerprints Twilio developer environments, exfiltrates host/process context and environment variables via webhook, later versions specifically target Twilio ACCOUNT_SID/AUTH_TOKEN and inject into node_modules, and some builds probe Twilio-related hosts plus AWS instance metadata. Not affiliated with Twilio; RL assesses it almost certainly violates Twilio HackerOne rules of engagement (no customer/employee data exfiltration). Package no longer present under that publisher as of RL write-up. Primary: ReversingLabs blog; wire: The Hacker News 22 Sep 2026. Correction (24 Oct 2026): an earlier version said the 11 versions were published over about 45 minutes; neither ReversingLabs nor The Hacker News gives that figure. ReversingLabs says they were published back to back on 14 August. (source: ReversingLabs)
ReversingLabs — malicious npm campaign targeting Twilio developers
breaches identity cloud
Incidents
Tue 22 Sep
Updated Fri 23 Oct
ShinyHunters (22 September 2026) claims it breached FBI systems using an alleged Oracle PeopleSoft remote-code-execution zero-day and stole employee and applicant data. The group shared a screenshot of apply.fbijobs.gov allegedly defaced with its Umbreon logo; the jobs/special-agent portals later showed maintenance. 404 Media reports receiving a sample of agent names, home addresses, and phone numbers (including spouses) and verifying a portion against public records; TechCrunch and BleepingComputer relay the claim. BleepingComputer states it has not independently verified the alleged zero-day, lateral movement, or volume of stolen data. UPDATE 26 Sep 2026: BleepingComputer reports the FBI confirmed it is investigating the claims without confirming a breach or data theft. Separate Google Mandiant/GTIG report the same week documents renewed UNC6240 mass exploitation of already-patched CVE-2026-35273 (PeopleSoft PeopleTools PSEMHUB) with WAF-bypass encoding and SIDEEYE — desk tracks that campaign on card cve-2026-35273; unclear if the FBI claim is that CVE or a distinct flaw. Primary wire: BleepingComputer; corroboration: 404 Media / TechCrunch. Switch primary_url if FBI or Oracle publishes a confirmed notice. CORRECTION (21 Oct 2026): we removed two later updates, a 30 September update quoting ShinyHunters' statements to the media and a 6 October update saying the FBI removed an Accenture contractor, because the stories they relied on were not linked or could not be read. They will return only with readable linked sources. Correction (23 Oct 2026): removed ShinyHunters' own claims about data volume, systems reached and other targets; BleepingComputer says it has not verified them. (source: https://www.bleepingcomputer.com/news/security/shinyhunters-claims-fbi-hack-data-theft-in-peoplesoft-zero-day-breach/)
BleepingComputer — ShinyHunters claims FBI / PeopleSoft breach (22 Sep 2026)
breaches identity cloud
Microsoft On the Issues / Digital Crimes Unit (22 September 2026) and Microsoft Threat Intelligence (“Unmasking EvilTokens”) detail disruption of EvilTokens, a phishing-as-a-service platform Microsoft tracks as Storm-2992. The service abused OAuth 2.0 device-code authentication against Microsoft’s legitimate sign-in flow (MFA-bypass style session/token theft without password capture) and centred an AI-style chatbot that analysed compromised mailboxes to pick BEC targets, impersonation paths, and fraud plays. Microsoft links >12,000 compromised inboxes across >10,000 organisations since the Feb 2026 launch; highest victim concentrations include the United States, Canada, the United Kingdom, Australia, India, and France. Partners helped seize ~50 operator sites and disable 150+ related domains; UK Metropolitan Police arrested two suspected administrators. Sold via Telegram (~US$1,500 init + US$500/month). Primary: Microsoft On the Issues; TI: aka.ms/EvilTokens-PhaaS; wire: BleepingComputer 22 Sep 2026.
Microsoft On the Issues — Disrupting EvilTokens (22 Sep 2026)
ai identity australia
Incidents
Mon 21 Sep
Updated Mon 26 Oct
DriveWealth says the incident was contained on 5 September 2026 (DriveWealth notice of data breach, filed with the California Attorney General; see Updated 30 Sep 2026 below). Stake, the Sydney-founded investing platform, told customers on 21 September 2026 that DriveWealth, the US broker that provides execution, custody and clearing for Stake Wall St accounts, had a cyber incident that affected some Stake customers' personal information. Stake says the information involved varies by person and may include name, email address, phone number and postal address; W-8 or W-9 tax status and country of taxation; the DriveWealth (Stake Wall St) account number; and aggregate snapshots of portfolio value, cash balance and buying power. It says Stake login credentials, tax file numbers and other government ID numbers, bank account details, identity documents, individual holdings and trading history were not accessed, that Stake's own systems were not affected, and that DriveWealth reported no unauthorised trading, transfers or withdrawals. People who opened a Stake Wall St account but no longer use it may also be affected. Stake is emailing affected customers. New Zealand's Hatch and the UK's Freetrade (for former Stake UK customers) have issued similar notices about the same DriveWealth incident. Stake has not said how many Australians are affected. Primary: Stake (21 Sep 2026); vendor: DriveWealth breach notice. Updated 25 Sep 2026 (Nine.com.au, 10:28 am Perth): Nine reported that Stake apologised to customers and said it was working with DriveWealth to understand the extent of the incident. Updated 30 Sep 2026 (DriveWealth; the California Attorney General's breach list gives 30 September, US date): DriveWealth's own breach notice says unauthorised access to its network took place on 4 and 5 September 2026, that it contained the compromise that day, and that its investigation and document review concluded on 28 September. Correction (26 Oct 2026): Nine.com.au's report and DriveWealth's notice, both later than Stake's 21 September statement, now carry Updated lines. (source: Stake; Nine.com.au; DriveWealth) Correction (26 Oct 2026): the card now leads with DriveWealth's statement, in its linked breach notice, that the incident was contained on 5 September 2026; no linked source gives a different date. (source: DriveWealth)
Stake — DriveWealth Data Security Incident (21 Sep 2026)
breaches australia identity
Incidents
Mon 21 Sep
Updated Mon 26 Oct
Qantas is investigating a possible security breach after some Qantas Hotels customers were targeted by WhatsApp phishing messages, Australian Aviation reported on 21 September 2026. Qantas said it was investigating reports of phishing attempts targeting some Qantas Hotels customers, that anyone who receives a suspicious WhatsApp message claiming to be from a hotel should ignore it, that Qantas Hotels and its accommodation partners will never ask customers to verify a booking or provide personal details via WhatsApp, and that customers concerned about a booking should check Qantas.com or the Qantas app; it also said it never asks customers for passwords. Capella Sydney said it was aware of reports from guests who had received suspicious WhatsApp messages requesting payment and credit card details in relation to their reservations, and confirmed the messages did not come from the hotel. According to Australian Aviation, the Australian Financial Review broke the news the previous week, and neither Qantas nor Capella Sydney had any further update. No threat actor has been identified, and neither company has said how the senders knew who had a booking. The campaign follows the 2025 breach in which more than five million Qantas customers' data was compromised. Primary: Australian Aviation (21 Sep 2026). Correction (26 Oct 2026): the card date has moved from 17 to 21 September 2026, the date of Australian Aviation's report; the Financial Review report the card was first dated by is not linked. Qantas's scams page cannot be read by this desk, so details only it gave (a September 2026 entry for the campaign and its mention of countdowns and cancellation threats) have been removed, along with a description of the messages quoting hotel names and booking dates, which Australian Aviation does not give. (source: Australian Aviation)
Australian Aviation — Qantas investigates WhatsApp phishing reports targeting customers (21 Sep 2026)
breaches australia
GreyNoise (Andrew Thompson, Mark Mager; 21 September 2026) tracks a suspected Chinese-speaking malicious cyber actor (“Kapibala” tooling) that from ~20 July 2026 exploited the WordPress wp2shell chain CVE-2026-63030 and CVE-2026-60137 against ≥49 organisations in 29 countries (SMB and government). In the most serious observed intrusion against a western government victim, the actor deployed a custom webshell/plugin, created a hidden admin, harvested credentials, pivoted to an internal SQL server, and downloaded ≥18,566 records including accounts, plaintext passwords, and PII tied to law enforcement/government agencies (timeline on 22 July 2026). GreyNoise also observed the same actor targeting Zyxel GS1900 switches (996 devices / 48 countries compromised or had sensitive data exfiltrated) — related desk context: cve-2026-7273 KEV card. Distinct from wordpress-click2shell-20260918 / Comment2Shell (different flaw family). Primary: GreyNoise blog; wire: Cyber Security News 22 Sep 2026.
GreyNoise — Open Season on Kapibala / WordPress wp2shell gov theft (21 Sep 2026)
breaches cloud network
