Threat Brief — 2026-07-16: AI Weapons, Supply Chain Squeeze

Executive summary: Today's intel is dominated by two converging trends: the weaponisation of AI tooling by threat actors and a fresh wave of supply-chain and developer-tool compromises. CISA is actively warning about exploited SharePoint flaws, while SonicWall SMA 1000 zero-days are under active attack. Meanwhile, malicious npm packages, Cursor IDE flaws, and a new Windows zero-day PoC underscore that the developer environment is now a primary battleground.

Top Items

• SonicWall SMA 1000 zero-days actively exploited — Two zero-days in Secure Mobile Access 1000 appliances, one enabling arbitrary command execution as admin. Patch immediately if exposed. (The Hacker News)
• CISA: Actively exploited SharePoint vulnerabilities — Three CVEs targeting internet-exposed on-prem SharePoint Server instances are being exploited in the wild; one (CVE-2026-58644KEV) had its exploitability flags corrected today. Prioritise patching. (BleepingComputer)
• Malicious AsyncAPI npm packages — Five infected @asyncapi namespace packages delivered a multi-stage botnet loader with credential-stealing RAT capabilities via npm supply-chain attack. Audit dependency manifests. (BleepingComputer / The Hacker News)
• Cursor IDE auto-executes git.exe from cloned repos on Windows — Opening a malicious repository in Cursor silently runs any git.exe in the project root with no prompt. Devs should avoid cloning untrusted repos and enforce workspace trust settings. (The Hacker News)
• New Windows zero-day PoC "LegacyHive" dropped post-Patch Tuesday — Researcher Chaotic Eclipse published a PoC for a User Profile Service arbitrary hive load EoP vulnerability. Patch status unknown; monitor for exploitation. (The Hacker News)
• Zoom critical account takeover vulnerability — Unauthenticated attacker could hijack accounts via the Windows desktop client/SDK. Update Zoom clients immediately. (BleepingComputer)
• Google Gemini CLI abused as hacking agent / botnet operator — Threat actor "bandcampro" leveraged the open-source Gemini CLI to automate attacks and operate a small botnet. Watch for misuse of AI CLI tools in your environment. (BleepingComputer)
• Identity attacks now top ransomware root cause — Email-based credential attacks overtook exploits as the primary ransomware vector; MFA (deployed in 97% of cases) failed to prevent compromise. Reinforce identity-layer controls beyond basic MFA. (Dark Reading)
• RabbitMQ CVE cluster — Microsoft published nine RabbitMQ vulnerabilities (CVE-2026-57211 through CVE-2026-44839) covering XSS, SSRF, credential disclosure, DoS, and cross-tenant bypass. If you run RabbitMQ management UI or federation plugins, review and patch. (MSRC)
• Secure Boot blind spot from forgotten/revoked UEFI shim bootloaders — Nearly a dozen revoked shims remained trusted for years, enabling Secure Boot bypass. Audit UEFI bootloaders across your fleet. (Dark Reading)
• OkoBot malware targets hardware wallet seed phrases — Windows malware framework active since April 2025 injects seed-phrase phishing into Ledger/Trezor app UI. Relevant for users handling crypto assets on managed endpoints. (The Hacker News)
• TuxBot v3: LLM-assisted IoT botnet framework — Unit42 analyses a new IoT botnet whose code shows LLM-assisted development, including cross-compiled binaries and C2 architecture. Low direct enterprise risk but signals lowering barrier to botnet development. (Unit42)
• Firefox/Chrome/Adobe/VMware critical patches — Mozilla warns of published exploit code for CVE-2026-15718 (WebAssembly invalid pointer) in Firefox; Chrome, Adobe, and VMware also shipped critical fixes. Patch browsers and endpoints. (The Hacker News)

Themes

• Developer environments under sustained attack: Malicious npm packages, a Cursor IDE auto-execution flaw, a "2-click Cursor exploit," and a new Windows zero-day PoC all surfaced within 24 hours. The dev toolchain is being actively targeted — enforce workspace trust, dependency scanning, and secret hygiene.
• AI tooling dual-use risk: Multiple findings show LLMs being used both offensively (Gemini CLI as hacking agent, TuxBot v3 LLM-assisted botnet, "vulnerability vending machine") and defensively. The frontier is shifting fast — monitor for AI CLI/tool abuse in your environment.
• Identity remains the perimeter: With identity attacks now the leading ransomware vector and MFA proving insufficient, expect renewed focus on adaptive authentication, session token protection, and email gateway hardening.
• RabbitMQ batch disclosure: Nine CVEs published in one day across RabbitMQ components (management UI XSS/SSRF, AMQP loopback bypass, credential disclosure). If RabbitMQ is in your stack, treat this as a coordinated patch window.