# internetonfire > A central, community-maintained tracker of recent cybersecurity and AI incidents: breaches, backdoored packages, exploited vulnerabilities and AI platform incidents, each linked to its original source and labelled by how solid the evidence is. Maintained by Kuber Mehta (@kuberwastaken). Entries are triaged by an AI agent and corrected by the community through GitHub. Last source check: 2026-10-22. Severity (how bad it is, judged on what happened rather than CVSS alone): **critical** (exploited or compromised at scale; secrets, data or code execution in attackers’ hands); **high** (real compromise with bounded scope, or a severe flaw that needs urgent patching); **medium** (patched with no known exploitation, contained, or limited impact); **low** (informational: a downstream response or an unsubstantiated claim). Evidence (how well-sourced it is): **confirmed** (a primary source backs it); **reported** (credible reporting, no primary confirmation); **disputed** (sources disagree); **unverified** (a claim going around, not established). Entries are grouped by the day they went public, not the day they happened. An outage is not proof of a breach; a CVE is not proof of exploitation. Quote the status label along with any claim. ## Data - [incidents.json](https://kuber.studio/internetonfire/data/incidents.json): the full log. Each incident has id, name, subject, summary, severity, status, date, disclosureDate, tags, sources (label, url, kind: primary|report|post|response, date), an optional correction and an optional detail object (whatHappened, exposure, fix, remaining, responses, claimChecks). - [tags.json](https://kuber.studio/internetonfire/data/tags.json): the tag vocabulary. - [Atom feed](https://kuber.studio/internetonfire/feed.xml): newest incidents. - Link to one incident with `https://kuber.studio/internetonfire/#`, or `https://kuber.studio/internetonfire/i//` for a link with its own preview card. ## Contribute Submissions are GitHub issues and pull requests on [Kuberwastaken/internetonfire](https://github.com/Kuberwastaken/internetonfire). Bring an original source. No credentials, leaked data or exploit code. ```sh # report an incident (opens the issue form) gh issue create -R Kuberwastaken/internetonfire --web # vote on a submission in review (+1 or -1) gh api -X POST repos/Kuberwastaken/internetonfire/issues//reactions -f content=+1 ``` To change the data directly, edit data/incidents.json and open a PR; the [contribution guide](https://github.com/Kuberwastaken/internetonfire/blob/main/CONTRIBUTING.md) has the format. Tags: breach, supply-chain, npm, pypi, ci-cd, malware, ransomware, vulnerability, exploited, rce, credentials, data-leak, cloud, ai, ai-assisted, pki, outage, state-actor, advisory, privacy, protocol, network, open-source, wordpress, healthcare, education, retail, finance, government, india. ## Incidents - [LiteLLM 1.107.0 shipped a credential stealer via PyPI and Docker](https://kuber.studio/internetonfire/i/litellm-supply-chain-v11070/): Critical severity, confirmed, disclosed 2026-10-21. LiteLLM confirms its release pipeline pushed malicious 1.107.0 packages to PyPI and container registries. The malware stole credentials and loaded remote-code tools. Affected artifacts were removed; private enterprise releases were not impacted. Tags: supply-chain, pypi, malware, credentials, ai. - [Onamae: malware planted on up to 26,739 Desktop Cloud servers](https://kuber.studio/internetonfire/i/onamae-desktop-cloud-malware/): Critical severity, confirmed, disclosed 2026-10-21. Onamae says malicious programs were placed in customer Desktop Cloud server environments. Up to 26,739 servers were potentially affected, and it began stopping servers to contain the incident. Data viewing was possible; data theft was not confirmed. Tags: breach, malware, cloud. - [PHP2Shell: two critical 9.2 RCE flaws in PHP core](https://kuber.studio/internetonfire/i/php2shell-cve-2026-109618-109619/): Critical severity, confirmed, disclosed 2026-10-21. PHP fixed two critical core bugs, CVE-2026-109618 and CVE-2026-109619, that can allow unauthenticated remote code execution. The PHP team rates both 9.2 and urges immediate updates. All web-serving PHP installations need review, with the highest demonstrated risk in Apache mod_php. Tags: vulnerability, rce, open-source. - [Railway customer secrets possibly decrypted; rotate everything](https://kuber.studio/internetonfire/i/railway-env-vars-breach/): Critical severity, confirmed, disclosed 2026-10-21. Railway says an unauthorised party may have accessed encrypted customer environment variables and decrypted some of them. It paused deployments during its investigation and recommended credential rotation. The affected-account scope remained unresolved. Tags: breach, cloud, credentials. - [Storm-0940 used agentic AI to attack hundreds of orgs, says Microsoft](https://kuber.studio/internetonfire/i/microsoft-artex-storm0940/): Critical severity, confirmed, disclosed 2026-10-21. Microsoft Threat Intelligence reports that China-state-aligned Storm-0940 used ARTEX agentic AI to automate attacks targeting hundreds of organizations. Some attacks gained privileged cloud access in under three hours. Microsoft attributes the ConnectWise supply-chain compromise to this campaign; that incident is also listed separately with its vendor response. Tags: ai, ai-assisted, state-actor. - [GitLab: CVSS 10.0 RCE affecting 12 years of releases](https://kuber.studio/internetonfire/i/gitlab-cve-2026-110264-onigmo/): High severity, confirmed, disclosed 2026-10-21. GitLab released fixes for CVE-2026-110264, a CVSS 10.0 flaw in its bundled regular-expression compiler that could allow unauthenticated code execution under certain conditions. Older self-managed releases are affected; no exploitation was known in the captured reporting. Tags: vulnerability, rce. - [333 Tor exit relays rerouted SSH to harvest passwords](https://kuber.studio/internetonfire/i/tor-exit-relay-hijack/): High severity, reported, disclosed 2026-10-21. A researcher found 333 Tor exit relays (~10% of the network's ~3,000 exits) redirecting SSH connections through 12 interception servers using the Ebury OpenSSH backdoor, harvesting passwords when users accepted forged host keys. Exit policies on hundreds of relays were rewritten Oct 1-5; Tor directory authorities removed the compromised relays around Oct 20-21. Tags: network, privacy. - [Bullhorn: contained cyber incident, then a separate login outage](https://kuber.studio/internetonfire/i/bullhorn-outages-october-2026/): Medium severity, confirmed, disclosed 2026-10-21. Bullhorn’s status updates describe separate events: a contained 17 October cyber incident, precautionary PHP2Shell product shutdowns, and a 21 October login outage linked to database-replication lag. They should not be presented as one confirmed breach. Tags: outage. - [Group claims Rajasthan Police breach and offers data for sale](https://kuber.studio/internetonfire/i/rajasthan-police-allegation/): Medium severity, unverified, disclosed 2026-10-21. A group calling itself Scarlettgroup claims to have breached Rajasthan Police and offered data for sale. Coverage repeats claims of about 70,000 identity records and 59,460 mobile numbers. These figures and the alleged breach are unverified. Tags: data-leak, government, india. - [Node.js patches a Brotli RCE and TLS hostname bypasses](https://kuber.studio/internetonfire/i/nodejs-october-2026-security-release/): Medium severity, confirmed, disclosed 2026-10-21. Node.js shipped October security releases covering four high-severity flaws alongside medium- and low-severity fixes. The release also updates certificate trust after eMudhra. Patched vulnerabilities do not by themselves establish an intrusion. Tags: vulnerability, open-source. - [SBI customer records surface on the dark web](https://kuber.studio/internetonfire/i/sbi-data-breach/): Medium severity, reported, disclosed 2026-10-21. SBI told stock exchanges that customer records had surfaced on the dark web, according to reporting. Reports distinguish more than 375,000 records from 3,897 containing personal information. The bank was still assessing authenticity and scope. Tags: data-leak, finance, india. - [Signal call bug could keep audio flowing after hang-up](https://kuber.studio/internetonfire/i/signal-ringrtc-hangup-cve-2026-113380/): Medium severity, reported, disclosed 2026-10-21. Reporting describes a RingRTC flaw that could let a malicious participant continue receiving audio after the other person hangs up. Camera continuation is not confirmed. The issue concerns call shutdown, not a general break of Signal encryption. Tags: vulnerability, privacy. - [Tailscale ships update distrusting eMudhra certificates](https://kuber.studio/internetonfire/i/tailscale-ts-2026-013-emudhra/): Low severity, confirmed, disclosed 2026-10-21. Tailscale published an emergency update to distrust the compromised eMudhra intermediate certificates. This is a downstream certificate-trust response, not a separate Tailscale breach. WireGuard mesh traffic does not rely on that public certificate trust. Tags: pki. - [Unclear North Korea campaign claim, no confirmed match](https://kuber.studio/internetonfire/i/north-korea-campaign/): Low severity, unverified, disclosed 2026-10-21. The original list does not identify which North Korea-linked campaign it means. Reporting on Operation FlashHole is one possible lead, not a confirmed match to that claim. Tags: state-actor. - [Plex2Shell: unauthenticated RCE in Plex Media Server, exploited in the wild](https://kuber.studio/internetonfire/i/plex2shell/): Critical severity, confirmed, disclosed 2026-10-20. Plex confirms active exploitation of a chain of server vulnerabilities that can let an unauthenticated attacker run code. Update to 1.43.3.10861 or newer; Plex says an older internet-reachable server may already be compromised. CISA lists the four related CVEs as known exploited. Tags: vulnerability, exploited, rce. - [CISA/FBI: SafePay ransomware has hit 600+ victims](https://kuber.studio/internetonfire/i/safepay-ransomware/): High severity, confirmed, disclosed 2026-10-20. A new CISA/FBI advisory describes SafePay ransomware activity and at least 600 victims across critical infrastructure sectors. The advisory is new, not the group: SafePay has operated since 2024. Tags: ransomware, advisory. - [Ransomware gangs exploit a patched PAN-OS GlobalProtect auth bypass](https://kuber.studio/internetonfire/i/paloalto-ransomware-attacks/): High severity, reported, disclosed 2026-10-20. Reporting links multiple ransomware groups to exploitation of a patched PAN-OS GlobalProtect authentication bypass. The attacks began months before this week’s reporting; this is not a newly discovered October intrusion. Tags: ransomware, exploited, vulnerability. - [Eight eMudhra CA keys were crackable from public certificates](https://kuber.studio/internetonfire/i/emudhra-ca-key-compromise/): High severity, confirmed, disclosed 2026-10-20. Eight eMudhra intermediate certificate-authority keys were weak enough to recover from public certificates, including one with the value 4. eMudhra revoked them. Its account says none issued end-entity certificates; this is not a confirmed breach of its infrastructure. Tags: pki, india. - [Paradigm contractor breach hits 970,000+ Medi-Cal members](https://kuber.studio/internetonfire/i/medical-paradigm-breach/): High severity, reported, disclosed 2026-10-20. Reporting says a Paradigm Healthcare Services breach affected more than 970,000 Medi-Cal members. The intrusion was detected in October 2025; notifications followed in September 2026. This is a contractor incident, not a confirmed breach of California’s own systems. Tags: data-leak, healthcare. - [HTTP/3 QPACK bug can exhaust server memory in 7 of 10 stacks](https://kuber.studio/internetonfire/i/http3-header-accordions-qpack/): Medium severity, confirmed, disclosed 2026-10-20. Foxio’s research describes QPACK header-decompression amplification that can exhaust server memory. Seven of ten tested implementations were affected, with patch status varying by implementation. This is a vulnerability report, not proof that all affected servers were attacked. Tags: vulnerability, protocol. - [TeamUpdraft breach: attackers added admins on connected WordPress sites](https://kuber.studio/internetonfire/i/teamupdraft-breach/): Critical severity, confirmed, disclosed 2026-10-19. TeamUpdraft confirmed unauthorised access over 17–18 October. It believes its customer database was accessed and says attackers abused UpdraftCentral connections to create admin accounts and install malicious plugins. The full extent is unknown. Connection keys were revoked; the plugin downloads themselves were not affected. Tags: breach, supply-chain, data-leak, wordpress. - [Cairn AI agents abused for intrusions; researcher claims 100+ companies hit](https://kuber.studio/internetonfire/i/cairn-ai-agents-100-companies/): High severity, reported, disclosed 2026-10-19. Norforge reports exposed dashboards showing threat actors abusing the open-source Cairn agent platform for intrusion attempts. Its researcher told press that one actor breached more than 100 companies. This is a research assessment, not a verified victim-by-victim count. Tags: ai, ai-assisted, open-source. - [OVHcloud CLI release binaries swapped for a credential stealer](https://kuber.studio/internetonfire/i/ovhcloud-cli-backdoor/): High severity, confirmed, disclosed 2026-10-19. OVHcloud says an attacker replaced official CLI release binaries with credential-stealing malware. Versions 0.11.0–0.15.0 were affected during a roughly 94-minute window on 19 October; 0.16.0 was not. Tags: supply-chain, malware, credentials, cloud. - [PingCAP: attackers copied TiDB Cloud data and source code](https://kuber.studio/internetonfire/i/tidb-cloud-breach/): High severity, disputed, disclosed 2026-10-19. PingCAP confirms unauthorised access to internal cloud resources between 9 September and 7 October, including copied data and source code. The breach is confirmed, but claims about plaintext keys and the full customer impact are disputed or unresolved. Tags: breach, cloud, data-leak. - [OpenCode breach exposes users' AI provider keys](https://kuber.studio/internetonfire/i/opencode-api-keys-distillation-attack/): High severity, confirmed, disclosed 2026-10-19. OpenCode confirmed a security breach affecting 1.4% of Console users during a large-scale attack. Reporting says stored keys for linked AI providers were exposed and some users saw unauthorised charges. The official post confirms the breach, not an audited total of financial losses. Tags: breach, ai, credentials. - [Signed Siemens Software Center client abused to side-load malware](https://kuber.studio/internetonfire/i/siemens-software-center/): High severity, confirmed, disclosed 2026-10-19. Attackers are abusing a DLL side-loading flaw in the signed Siemens Software Center client through ClickFix lures. Siemens says its systems and update channel were not breached. Tags: vulnerability, exploited, malware. - [Ahold Delhaize cyberattack knocks out websites, apps and deliveries](https://kuber.studio/internetonfire/i/ahold-delhaize-cyberattack/): High severity, reported, disclosed 2026-10-19. Ahold Delhaize took systems offline after detecting malicious activity, disrupting websites, apps and deliveries. Stores remained open. Its 19 October update reported recovery progress and no evidence then of sensitive data theft. Tags: outage, retail. - [ConnectWise backup credentials used to sign into 300+ Microsoft 365 tenants](https://kuber.studio/internetonfire/i/connectwise-cloud-backup/): Critical severity, confirmed, disclosed 2026-10-16. Compromised ConnectWise Cloud Backup application credentials enabled unauthorized Microsoft 365 sign-ins. Huntress saw sign-in activity in almost 1,000 tenants, over 300 successful; these are sign-in counts, not proven data-theft counts. Affected tenants need application-credential and audit review. Tags: breach, credentials, cloud. - [ETS breach exposes three school years of California student records](https://kuber.studio/internetonfire/i/california-ets-student-data-breach/): High severity, confirmed, disclosed 2026-10-16. California’s education department notified families about files copied from testing contractor ETS. The records span three school years and include sensitive student information. The copying happened between May and September, not during this week’s reporting. Tags: data-leak, education. - [Rank Math PRO 3.0.123 shipped with a self-reinstalling backdoor](https://kuber.studio/internetonfire/i/rankmath-pro-malware/): High severity, confirmed, disclosed 2026-10-16. The official Rank Math PRO 3.0.123 download contained malware during an 11-hour window on 15–16 October. The vendor released clean version 3.0.124 and published recovery guidance. The free WordPress plugin was not affected. Tags: supply-chain, malware, wordpress. - [Zeabur breach: 3.2M environment variables read; platform shutting down](https://kuber.studio/internetonfire/i/zeabur-breach/): Critical severity, confirmed, disclosed 2026-10-15. Zeabur confirms attackers read 3,218,965 environment-variable records and data for 321,810 accounts, among other affected systems. It revoked credentials and urges secret rotation and migration. Some services begin shutting down Dec 1; all dedicated servers and shared clusters stop by Dec 31, 2026. Tags: breach, cloud, credentials. - [AOMEI Backupper updates hijacked to deliver malware to selected users](https://kuber.studio/internetonfire/i/aomei-update-hijack/): Medium severity, confirmed, disclosed 2026-10-15. Google’s threat researchers found brief tampering with AOMEI Backupper update infrastructure that redirected selected users to malware. The attacks happened between April and July; the research was published in October. Tags: supply-chain, malware. - [NGINX patches five medium-severity bugs](https://kuber.studio/internetonfire/i/nginx-october-2026-advisories/): Medium severity, confirmed, disclosed 2026-10-14. NGINX’s October release fixed five newly published security advisories. The vendor rates all five medium, not critical. One concerns HTTP/3; this is distinct from the separate Header Accordions research. Tags: vulnerability, open-source. - [Tensorlake npm release carried a Shai-Hulud-style worm](https://kuber.studio/internetonfire/i/tensorlake-npm-worm/): High severity, confirmed, disclosed 2026-10-08. tensorlake@0.5.144 was published to npm on Oct 8, 2026 carrying a Shai-Hulud-style credential-stealing worm with a destructive dead-man's switch ('hostage token'). The Register framed it as the Shai-Hulud worm's jump into AI infrastructure. Tags: supply-chain, npm, malware, credentials, ai. - [GhostAction: malicious GitHub workflows target 30,000+ repos for secrets](https://kuber.studio/internetonfire/i/ghostaction-github/): High severity, confirmed, disclosed 2026-10-07. GhostAction injected credential-stealing GitHub Actions workflows through compromised accounts. The October wave targeted over 30,000 repositories and attempted to extract 3,174 secrets. Those target/attempt counts are not confirmed theft totals; the earlier August-September wave had 26 confirmed stolen secrets. Tags: supply-chain, ci-cd, credentials. - [@subql/common 5.8.3 on npm steals credentials on install or import](https://kuber.studio/internetonfire/i/subquery-npm-backdoor/): High severity, confirmed, disclosed 2026-10-05. The @subql/common 5.8.3 npm release contained a credential-stealing payload, according to a public security report and maintainer response. It was published through the project’s release pipeline with npm provenance; provenance alone did not make it safe. Tags: supply-chain, npm, malware, credentials.