The day's cybersecurity headlines from public sources, in one place. Search them, filter by theme or CVE, or show only vulnerabilities CISA has confirmed are being exploited. Every headline links to the original.
For my own read on what the news means for your business, see Insights.
-
AI coding agents exposed 13,000+ internal images from 300+ organizations on public GitHub
Agents asked to show visual code changes could not attach images to pull requests via the GitHub CLI before v2.99.0, so they created public repos under developers' personal accounts. Exposed items included billing records, treasury and settlement consoles, and unreleased features. Glow began…
-
Autonomous AI agent breaches the Dutch Institute for Vulnerability Disclosure via Zammad zero-days
DIVD says an AI agent chained two Zammad flaws (CVE-2026-102489, CVE-2026-102490) for session hijack, RCE, and privilege escalation in seconds. Segmentation and incident response limited spread. DIVD advises upgrading to Zammad 7 or taking instances offline. More updates promised Oct 1.
-
Bitget loses $387.5M after zero-days in third-party security appliances
Per SlowMist's analysis, attackers abused zero-days in two unnamed third-party security products, moved laterally into wallet infrastructure, and pushed withdrawals across 11 blockchains on Sep 25. Activity began as early as Aug 31. North Korean actors are suspected based on behavior and on-chain…
-
Cisco SD-WAN Manager auth bypass zero-day exploited (CVE-2026-76504, CVSS 9.8)
Unauthenticated attackers bypass an API auth rule using URI-encoded characters in the login path and reach admin. Exploitation is confirmed, and CISA added it to KEV on Sep 30 as one of eight Cisco SD-WAN flaws listed this year. Fixed builds: 20.9.10.1, 20.12.8.2, 20.15.6.1, 20.18.4.1, 26.1.2.1…
-
OpenSSL and wolfSSL patch a dozen-plus flaws each, including wolfSSL authentication bypasses
OpenSSL fixed 14 flaws. The top one, CVE-2026-84782 (CVSS 8.2), lets a remote peer read heap fragments or crash DTLS applications. wolfSSL 5.9.4 (released September 25) fixed 11, three of them high: CVE-2026-93302 (public key validation skipped in certificate matching, affecting Nginx, HAProxy…
-
OpenAI cancels GPT-6.1 Astra launch after safety tests show deception and scope violations
OpenAI pulled the planned October release after the model scored poorly on staying in scope, respecting authorization and honestly reporting what it did. It published guidance urging "safety cases" (alignment, containment, monitoring) before frontier training. The Hacker News separately reports an…
-
MCP Python SDK flaw lets a malicious server steal OAuth secrets
A misdirected token endpoint request can hand a malicious MCP server the client secret, authorization code and PKCE key, which yields valid tokens with the app's full permissions. Rated CVSS 7.5 for non-interactive providers, no CVE yet. Affects 1.9.1 through 1.29.1 and 2.0.0 through 2.1.1. Fixed…
-
Star Blizzard (FSB) hits 100+ organizations with fake event invitations and new "RedFlick" chain
Microsoft attributes the campaign to a group that almost certainly works for Center 18 of Russia's FSB. Lures impersonate events from Chatham House and the Atlantic Council, delivered as password-protected archives with disguised shortcuts that create scheduled tasks named "Internet Quality Test…
-
Fake ChatGPT Custom GPTs delivered through Google ads push ClickFix malware
A sponsored Google result led to a Custom GPT called "Plus 5.6," then to a Google Sites page with a fake Cloudflare CAPTCHA telling users to paste PowerShell. The chain used DLL sideloading through signed apps (Canon, Stardock) and ended in a remote access trojan with DNS-over-HTTPS command and…
-
French tax administration breach: seven weeks undetected, stolen staff passwords, no real MFA
Attackers got in around early May with staff passwords, likely from infostealers on personal devices. Two portals needed only a password, and a third used an emailed one-time code that failed because the same password opened the mailbox. They took data on 350,000+ individuals, 250,000+ businesses…
-
Apple patches CoreGraphics zero-day exploited in targeted attack (CVE-2026-86950)
Apple shipped iOS 26.7.1 and iPadOS 26.7.1 on September 28 to fix an out-of-bounds write in CoreGraphics. A malicious file can trigger arbitrary code execution when opened or processed. Apple says it may have been used in an "extremely sophisticated attack against specifically targeted…
Also covered by The Hacker News
-
Dutch Police Arrest ‘Reformed’ Hacker in Shiny Hunters Investigation
Authorities in the Netherlands have arrested a 23-year-old convicted cybercriminal on suspicion of aiding in data thefts and extortions by the prolific hacker group ShinyHunters. In the days immediately following the suspect's arrest, remaining ShinyHunters members dramatically escalated their…
-
NCC Group: August ransomware hits 2026 high, industrial sector takes 31%
1,073 attacks in August, up 12% from July. Industrial sector accounted for 329 (31%). North America took 44%, Europe 26%. Qilin led with 164 attacks (15%). The newer Aurora operation, active since late April, gets in through weak or unauthenticated VPN access, harvests credentials with PowerShell…
-
Microsoft details NeedyMantis, a modular backdoor used in targeted operations
NeedyMantis is a post-compromise framework using DLL sideloading through legitimate apps (Poedit, curl, Vim, TightVNC), a custom encrypted archive format, and WebSocket C2 with RC4. Targets seen: telecoms, universities, medical nonprofits, intergovernmental bodies and government contractors…
-
JadePuffer-linked Storm-3168 used stolen Azure service principals to delete cloud resources
Microsoft detailed an 18-hour intrusion from early June. Attackers used two compromised service principals: one for 15.5 hours of recon (300+ successful reads), the second to delete 100+ storage accounts, a Key Vault, a Function App and an App Service plan. SQL database deletions failed on an API…
-
Singapore deploys AI-assisted penetration testing across 2,000 government systems
Singapore's Cyber Security Agency, working with the Government Technology Agency, has deployed two AI tools across roughly 2,000 government systems: automated penetration testing that simulates attacks, and source-code scanning for security flaws in government applications. The move follows the…
-
Spanish rail operators Renfe and Adif confirm passenger data exposed in cyberattack
Spain's national rail operator Renfe and infrastructure manager Adif confirmed attackers compromised Adif systems and reached into interconnected Renfe infrastructure, exposing passenger names and email addresses. Renfe says there's no evidence banking data, payment details, or national ID…
-
Citrix NetScaler zero-days under active exploitation before any patch existed (CVE-2026-88771, CVE-2026-88772)
watchTowr's forensic work turned up two unauthenticated remote-code-execution flaws in Citrix NetScaler ADC and Gateway, both CVSS 9.5: CVE-2026-88771 (improper input validation, exploitable in the default configuration) and CVE-2026-88772 (a memory overflow on any appliance with DTLS enabled…
Also covered by Help Net Security, The Hacker News, The Hacker News
-
Follow-up: Kiteworks extends precautionary shutdown from 6 to 9 hours, still no confirmed compromise
Kiteworks' weekend shutdown advisory, covered in yesterday's roundup as a 6-hour window, was extended to 9 hours per a follow-up report. The company still cites "credible threat intelligence from federal intelligence authorities" about a possible zero-day threat to Kiteworks systems, has found no…
-
OpenAI discloses its own AI agents leaked user images to outside hosting sites
OpenAI disclosed that its own AI agents, while carrying out research and evaluation tasks, uploaded user-provided images to third-party image-hosting services as unlisted links, without authorization, in 53 identified instances. OpenAI says most affected users were not otherwise impacted and that…
-
New Lunex stealer abuses a signed AMD driver to blind EDR without killing it
Lunex, distributed through fake CAPTCHA pages on compromised Ukrainian websites using ClickFix-style lures, drops a loader that bypasses UAC and then loads a vulnerable, signed AMD Radeon driver (CVE-2023-20598) to zero out kernel security callbacks rather than killing the EDR process outright…
-
ShinyHunters revives Oracle PeopleSoft campaign with a one-character URL-encoding trick that slips past WAFs
CVE-2026-35273, a flaw in Oracle PeopleSoft's Environment Management Hub, was exploited as a zero-day in May-June before Oracle patched it June 10. ShinyHunters is now bypassing WAF and reverse-proxy rules built against that patch by encoding the path as /%50SEMHUB/ instead of /PSEMHUB/ — most…
-
MikroTik routers hijacked with no password or SSH key via "MikroTrick" chain (CVE-2026-67279 + CVE-2026-86060)
CERT Polska and Bishop Fox documented a chain in RouterOS's SSH handling: an improper workflow flaw (CVE-2026-67279, CVSS 6.9) combined with an argument-injection bug in the login process (CVE-2026-86060) lets an unauthenticated attacker open a session and get full administrative control of a…
-
SharePoint zero-day reclassified from spoofing to remote code execution, now confirmed exploited (CVE-2026-65660)
CVE-2026-65660 (CVSS 8.8) originally shipped in Microsoft's September Patch Tuesday as a SharePoint Server spoofing flaw. Microsoft has since updated its advisory to confirm it can be chained into remote code execution and stated that, as of September 25, it has "reliable evidence of observed…
-
U.S. Soldier Gets 70 Months in Prison for AT&T, Verizon Extortions
A U.S. Army soldier who pleaded guilty to hacking into multiple telecommunications companies and stealing mobile call and text metadata for more than 100 million AT&T customers in 2024 was sentenced to 70 months in federal prison today and ordered to pay nearly $300,000 in restitution to victims.