Nitori warns customer records may have left its systems
Nitori reports unauthorized access and up to 37.62 million potentially exposed records. Passwords and card data are outside the reported scope; targeted scams remain a risk.
Nitori reports unauthorized access and up to 37.62 million potentially exposed records. Passwords and card data are outside the reported scope; targeted scams remain a risk.
A researcher found a credential embedded in the Android app that opened a Salesforce database. Wyndham revoked it, but has not explained past access or token invalidation.
The library says branches are open and borrowing works, but renewals and most public computers remain unavailable. Check the live service notice before visiting.
Stolen vendor credentials opened old cross-account roles. Check Qubole trust, inspect role sessions and S3 reads, and revoke exposed secrets.
Notion says fewer than 1% of workspaces may be affected. Owners can review the transcripts; rotate credentials pasted into affected chats.
Vates reports active attacks against its management server. Update both release channels, restrict management access and review affected hosts.
The CI runner provider says it contacted affected customers after an intruder accessed its job-log store. The investigation is ongoing.
An outside cloud service held records for roughly 7.7 million people. The retailer has confirmed unauthorized access, but is still investigating how many records were taken.
Apiiro found credentials in public agent settings and repository history. Review saved approvals, rotate exposed credentials, and check what travels with a project.
A researcher captured unencrypted authentication and reused it through the Windows launcher. Players need trusted connections; developers need encrypted transport and server-side session revocation.
Security analysis now runs by default, with wider tracking through PHP code and library models. Review newly visible findings and plugin compatibility before accepting a fresh baseline.
Microsoft now requires Paint 11.2609.401.0 to comprehensively fix CVE-2026-83844. Its Store-delivered update needs a separate version check, while the disclosed attack route remains thinly explained.
A trusted VS Code workspace ran malware on an unprotected Mac. A separate app authorization survived the password reset, showing why device health and grant revocation need explicit checks.
A developer says ChatGPT sent a confirmation code through the wrong connector. Their server rejected it; cross-provider leakage remains a hypothesis, while client-bound approval is a concrete defense.
Doctor Web found malicious build hooks and scripts in projects aimed at security practitioners. Treat repository setup as code execution, and investigate stolen sessions if a suspect project ran.
A new Security tab exposes per-file attestations and warnings about changed or missing origins. Inspect the file you will install; full provenance is not a code-safety verdict.
Valid credentials for an alternate login method can bypass SAML controls on affected SonicOS VPNs. CISA lists the flaw as exploited; Gen7 and Gen8 have separate firmware fixes.
Some Linux backup appliances panic during an early cryptographic check. Veeam's workaround changes reboot handling while a separate kernel security update still needs to reach the running system.
The compression library's release repairs conditional memory corruption and archive extraction flaws. Check the code paths and shipped copies, including minizip, and retain the warning about unsafe iostream2 input.
D-Link's interim advisory identifies tested DGS-3130-30TS B1 firmware and immediate discovery-protocol mitigations. Other models remain under investigation, and disabling DDP does not repair separate firmware weaknesses.
CERT-In documents embedded service credentials, exposed session tokens and weak network checks in iRA apps. Tata says fixes reached active users and its assessment found no evidence of unauthorized access.
Era observed attackers changing executable code per request on deliberately exposed Flowise honeypots. Saved workflows miss that activity; workload permissions determine how far a compromise can spread.
Truffle Security is coordinating issuer-side revocation across public code, images and datasets. Its reported key totals measure exposure, while operators still need replacement and recovery plans.
Push Security traced Microsoft device-code approvals and Google sign-in proxies behind fake AI events. Blocking unused login flows and revoking sessions address different parts of the attack.
A network migration exposed an internal system whose self-registered accounts could read archived customer data. DigiCert says certificate issuance and customer logins were unaffected.
XLab found a root-run fake updater that swaps CWP's management server and suppresses backdoor request logs. Initial access remains unknown; exposed panels are not a victim count.
Key-only accounts can become reachable when password login stays enabled and another account has a password. Alpine lists branch-specific repairs and cloud-image exposure.
CVE-2026-93389 affects plugin releases through 7.9.1 when REST caching serves personalized carts publicly. Upgrade, purge stored responses and review possible exposure.
Bitdefender demonstrated a permission-free app that repeatedly crashes a core Android service. Backups and installation controls matter while no verified fix is available.
Two August intrusions used Claude Code and ChatGPT desktop as execution parents. Elastic's new report makes process behavior more useful than an approved-app label.
The gradual vercel.com rollout protects session renewal with a device-held key. It leaves local malware, unsupported browsers and already-stolen credentials as separate problems.
June's firmware still allowed the recovered exploit. Zyxel's model-specific Patch 3 releases repair CVE-2026-107441; previously exposed switches also need investigation.
Customers report stolen encrypted cloud data; mSeven's initial notice leaves the scope unclear. Replacing account credentials and authentication secrets is separate from changing a vault password.
GMS 9.5.2 and earlier expose internal files without a login and database access with high privileges. SonicWall offers no fix or workaround for either deployment type.
Rails 8.0.5.2 and 8.1.4.1 repair Active Storage file access and other flaws. ImageMagick policies, media-tool builds and storage configuration decide the exposure.
Debian fixes a redirect-and-retry verification bypass in APT. Trixie needs 3.0.4 and Bookworm needs 2.6.2; distribution-specific updates matter.
Expo could sign attacker-supplied values; Electron could disclose server secrets. Update matching packages, rotate affected secrets and check session revocation.
Sekoia reports already-patched iOS and Android exploit chains delivered through websites and an analytics tracker. Browser and operating-system updates address different layers.
Bunny now reports completed certificate revocations in Let's Encrypt's lists. Its supplier investigation remains open; customer-supplied keys need separate confirmation.
Mandiant reports unauthorized model use costing some victims nearly $1 million. Restrict service activation, monitor new credentials and distinguish spending alerts from enforced limits.
CISA confirms active exploitation of CVE-2026-119977. The vulnerable window occurs during full core upgrades; fixed releases and forensic triage address different risks.
The Linux security suite distributed a backdoor through its update channel. Check update history, install the corrective release and invalidate exposed secrets.
Bitdefender's investigation traces credential theft, rogue administrators and cross-domain access. Suspected AI assistance remains an interpretation of the logs.
The provider revoked personal API tokens. Customers must also replace connected-service secrets and re-authenticate GitBook's MCP server.
Its new guide turns security basics into specific actions. Start with critical accounts, a tested backup and an offline response plan.
Two flaws can combine into root access on Email Security Gateway. Verify product-specific patches and restrict the management interface.
The provider confirms access to user information across internal systems. Account-specific notices matter more than guesses about the total breach size.
Its agent-assisted backports address a maintenance backlog. Human review, regression testing and distribution delivery still decide when users receive a fix.
WorkOS now identifies a public API, predictable development credential and missing rate limits. A researcher accepts responsibility; the claimed sandbox escape remains technically unexplained.
US agencies warn of ongoing attacks on cold-chain controls and add safe setpoint limits, logging and recovery guidance. The warning supplies no common exploit or new victim count.