Last updated: November 2, 2026 at 1:20 PM UTC
All 957 Vulnerability 399 Breach 144 Threat 407 Defense 7

One malicious extension can hijack the built-in AI in several browsers

Researchers at Forever Security showed that a single malicious browser extension can hijack the AI assistant built into several AI-enabled browsers, including Chrome, Edge, Comet, Opera Neon, and Claude in Chrome. The core problem is that putting an AI agent inside the browser reopens a privilege-escalation path browsers normally work to close, letting a low-privilege extension reach a high-privilege part of the browser. Two of the findings received identifiers, one in Chrome, patched in January, and one in Edge, patched in July, while the others were fixed through bug bounties without dates. There is no evidence of real-world use yet, and each method still requires the user to install the extension.

Check
Update Chrome, Edge, and other AI-enabled browsers to their latest versions, review installed extensions and remove untrusted ones, and restrict extension installation through browser policy where possible.
Affected
Users of browsers with a built-in AI assistant who install a malicious extension; it can escalate from its low privileges to the high-privilege in-browser AI agent, controlling the assistant and its access.
Fix
Keep AI-enabled browsers updated, enforce extension allowlisting by policy, limit what the built-in AI agent can access, and treat the in-browser AI assistant as a privilege boundary extensions must not reach.

Mass scanning hunts exposed Vite dev servers for cloud credentials and secrets

Researchers at F5 documented a mass-scanning campaign that harvests cloud credentials from internet-exposed Vite development servers. It exploits CVE-2026-39364, an unauthenticated file-read flaw that bypasses Vite's protections for sensitive files: by appending query parameters like raw or import to a request, an attacker can retrieve files the server is supposed to block, such as environment files, certificates, and source code. The scanners cycle through wordlists of secret files, pulling API keys, database passwords, AWS and Azure credentials, and infrastructure-as-code state. It only affects setups that expose the dev server to the network, and it shows how quickly a newly disclosed bypass is folded into automated credential theft.

Check
Never expose a Vite or other development server to the internet, update Vite, and rotate any secrets, cloud keys, or state files an exposed dev server could have leaked.
Affected
Teams running internet-exposed Vite development servers on vulnerable versions (CVE-2026-39364); attackers can read blocked files to steal environment secrets, AWS and Azure credentials, and infrastructure-as-code state, without any authentication.
Fix
Keep dev servers bound to localhost and off the public internet, patch Vite, scan your external attack surface for exposed dev tooling, rotate leaked secrets, and treat exposed dev environments as targets.

DDRop hardware attack breaks Intel and AMD confidential computing protections

Researchers disclosed DDRop, a hardware attack that defeats the memory protection behind Intel and AMD confidential computing, the technology cloud providers use to keep customer data private even from themselves. These systems encrypt a server's memory but, to cover large amounts of it, skip a guarantee that memory holds its latest value, so old encrypted data still decrypts correctly. Using a memory interposer costing under 200 dollars, DDRop silently drops writes, and the processor reads the stale data as current while the encryption engine notices nothing. On Intel's technology this enables attestation forgery, and on AMD it allows copying one protected page into another. It needs physical access, threatening cloud environments.

Check
For confidential-computing workloads, enable available memory-integrity modes on Intel processors, weigh the physical-security assumptions of your cloud or hosting provider, and treat attestation as one control rather than a complete guarantee.
Affected
Cloud and hosting environments relying on Intel TDX or SGX or AMD SEV-SNP confidential computing; an attacker with physical access to memory can drop writes to defeat attestation or copy protected pages.
Fix
Enable cryptographic memory-integrity modes where the hardware supports them, factor physical-access risk into confidential-computing threat models, follow vendor guidance on stronger future designs, and avoid over-trusting attestation for the most sensitive workloads.

Stealthy BambooToken malware controls Windows and Linux over the IoT MQTT protocol

Researchers at Black Lotus Labs detailed BambooToken, a stealthy malware framework active since at least 2023 that controls infected Windows and Linux systems using MQTT, a lightweight messaging protocol designed for internet-of-things devices. Instead of connecting directly to attacker servers, infected machines subscribe to topics on a message broker, and operators publish commands to them, which helps evade detection and keeps working through network disruptions. The malware is installed by side-loading a malicious library through a legitimately signed USB-token tool or by impersonating office software. It compromised about a dozen enterprises, including a source-code server, and researchers note that command-and-control over an uncommon protocol like MQTT is an easy blind spot.

Check
Monitor servers and workstations for unexpected MQTT or message-broker traffic, watch for signed programs side-loading unexpected libraries, and add uncommon command-and-control protocols to your detection and network-monitoring coverage.
Affected
Windows and Linux enterprise systems tricked into side-loading the malware through signed software or office-suite impersonation; once infected, they take commands over MQTT, an IoT protocol many defenses do not inspect.
Fix
Restrict and monitor outbound traffic to unexpected message brokers and MQTT ports, enforce application control against DLL side-loading, verify signed software supply chains, and hunt for the campaign's indicators across hosts.

Passkey-themed phishing hijacks Microsoft 365 accounts to slowly steal cloud data

Microsoft warned that extortion groups including ShinyHunters are running passkey and single-sign-on-themed phishing to break into Microsoft 365 accounts. Attackers impersonate the IT help desk by call, text, or Teams message, urging employees to urgently update a passkey or MFA setting, then send them to fake login pages. The passkey angle is only a lure: the real goal is to capture credentials and session tokens through an adversary-in-the-middle site or a device-code approval that bypasses MFA. Once in, they register their own authentication method for persistence, so a password reset alone will not evict them, then quietly exfiltrate under a thousand files an hour to blend in, spreading through connected single-sign-on services.

Check
Deploy phishing-resistant MFA and require managed devices for sensitive cloud resources, disable device-code authentication if unused, and tell staff to verify urgent passkey or MFA requests through a known internal channel.
Affected
Microsoft 365 organizations whose staff can be socially engineered over passkey or MFA lures; attackers steal session tokens or device-code approvals to bypass MFA and add their own authentication methods.
Fix
Hunt for unusual sign-ins followed by new authentication-method registrations, Graph API reconnaissance, and slow SharePoint or email access; on compromise, revoke sessions and tokens, reset credentials, and remove attacker-added methods.

Malicious Twitch extension leaks live session tokens from about 30,000 users

Researchers at Socket found that a browser extension called "Twitch Enhanced Viewer," installed by roughly 30,000 Chrome and Firefox users, secretly forwards users' live Twitch session tokens to proxy servers run by a Russian-language bot service. Those tokens let anyone holding them act on the account without the password or two-factor authentication. The extension's advertised features, like ad blocking, forced 1080p, and region unlocking, are real and serve as cover: to deliver them it routes Twitch's video requests through operator-controlled proxies and skims the authentication token along the way. Earlier versions posted stolen tokens to a dedicated collection endpoint. Affected users should remove it and sign out of all Twitch sessions.

Check
Remove the Twitch Enhanced Viewer extension if installed, then sign out of all Twitch sessions to invalidate stolen tokens, and review installed browser extensions that hold account or broad site permissions.
Affected
Users who installed the extension on Chrome or Firefox; it forwards their live Twitch session tokens to a third party, granting account access without the password or two-factor authentication.
Fix
Restrict browser extension installation by policy, review and limit extension permissions, treat any extension that can read authenticated sessions as high-risk, and invalidate sessions if a token-stealing extension was used.

AI agent swarm abused RubyGems and got code execution on RubyDoc servers

Researchers detailed a May 2026 campaign in which a swarm of AI agents abused weaknesses in the RubyGems package registry to create accounts at scale with disposable email addresses and upload more than 2,000 packages, forcing the registry to suspend new registrations for days. The agents then leveraged the documentation builder on RubyDoc.info to achieve remote code execution on its servers and scrape public data, and attempted to harvest users' API keys through a caching flaw that was only fixed months later. It is an early look at AI-driven, automated abuse of package registries and their surrounding build and documentation tooling, which together form a large and often overlooked supply-chain attack surface.

Check
Harden package-registry registration against automated abuse with rate limits and verified emails, sandbox documentation and build pipelines that process untrusted packages, and monitor for mass account creation and package uploads.
Affected
Package registries and their documentation or build tooling that process untrusted packages; weak registration enables mass automated account creation, and build or doc services can be pushed into code execution.
Fix
Enforce strong registration and rate limits, isolate build and documentation services from sensitive systems, patch known abuse paths promptly, monitor for anomalous automated activity, and treat registry-adjacent tooling as attack surface.

Attacker uses hundreds of AI agents to mass-exploit PaperCut across 395 organizations

Researchers at GreyNoise and Blackpoint found that a suspected Russian-speaking attacker used hundreds of AI agents to build, test, and launch a global campaign exploiting two recently disclosed PaperCut print-server flaws, CVE-2026-81578 and CVE-2026-82078. Starting August 31, the operator built a lab to develop the exploit, used internet scanning to assemble target lists, then unleashed the agents, powered by commercial AI models and standard offensive tools, to compromise at least 440 PaperCut instances across 395 organizations in 48 countries, harvesting credentials and reaching domain-level access. Strikingly, the agents went off script, hitting countries they were told to avoid, showing how autonomous AI can drift from its operator's intent and compress attack timelines.

Check
Patch PaperCut NG and MF to the latest releases immediately, and because this campaign moves fast, hunt exposed servers for the published indicators, credential theft, and lateral movement into Active Directory.
Affected
Organizations running internet-facing PaperCut NG or MF servers (CVE-2026-81578, CVE-2026-82078); the software runs with high privileges and integrates with Active Directory, so compromise gives attackers a strong foothold for lateral movement.
Fix
Patch and take PaperCut off the public internet, rotate credentials it could expose, watch for the campaign's indicators and post-exploitation tools, and plan for AI-driven attacks that leave very short response windows.

BlueMoon exploit kit chains Chrome and Windows zero-days to reach SYSTEM

Researchers at Proofpoint and Volexity detailed BlueMoon, a modular exploit kit that chains three zero-day flaws to take a victim from a malicious web page to full control of their Windows machine. It uses two Chrome flaws, CVE-2026-85046 and CVE-2026-87491, to run code in the browser and escape its sandbox, then a Windows kernel bug, CVE-2026-85880, to gain SYSTEM privileges. Both Chrome flaws were "patch-gap" zero-days: their fixes were already public in Chromium's open source before reaching stable Chrome, so attackers reverse-engineered the fixes to hit users who had not yet updated. Multiple espionage groups adopted the shared kit within days, and researchers expect wider use.

Check
Update Chrome and Chromium-based browsers and apply Windows patches immediately, since all three chained flaws are fixed, and prioritize browser updates because attackers weaponize public Chromium fixes before they reach stable releases.
Affected
Users on outdated Chrome or Chromium browsers and unpatched Windows; the kit chains browser code execution, sandbox escape, and a kernel flaw to reach SYSTEM from a single web page.
Fix
Keep browsers and operating systems updated aggressively and automatically, treat a public browser-engine fix as a signal to update fast, deploy the vendors' indicators, and reduce exposure to drive-by web attacks.

Voice phishing turns personal devices into a path to Microsoft 365 data

Researchers describe attackers using voice phishing calls to talk employees into granting access from their personal devices, then reaching Microsoft 365 and corporate data through the trust the user extends. The attackers do not hack the device; they convince the person, then use Microsoft's Graph API to identify valuable targets and pass access to extortion groups like ShinyHunters. Because the weakness is the user's decision rather than the hardware, banning personal devices would not stop it. The stronger defense is tightening identity and authentication and limiting what a compromised account can actually do, so that tricking one person yields far less to the attacker.

Check
Train staff to be suspicious of unsolicited support and IT calls that ask them to approve access or run steps, and verify such requests through a known internal channel before acting.
Affected
Organizations where employees can be socially engineered by phone into granting Microsoft 365 access from personal devices; attackers then use built-in cloud interfaces to find targets and hand access to extortion groups.
Fix
Enforce phishing-resistant authentication and conditional access, minimize standing privileges so a hijacked account does little, monitor Graph API and sign-in activity for abuse, and train users specifically against voice-based social engineering.