Monthly Cyber Threat Roundup

    Cyber Attacks & Data Breaches: May 2026

    A month-end digest of ransomware, data breaches, zero-days, supply chain compromises, state-linked activity and a dedicated look at AI-driven attacks.

    +48%ransomware surge in May (Check Point)
    275Musers claimed in the Canvas (Instructure) breach
    3,800GitHub repos stolen via a poisoned VS Code extension
    1stAI-developed zero-day exploit spotted by Google

    Read this first: this page is compiled from public reports and trackers. Counts differ between sources (leak-site claims vs confirmed attacks), attributions are often preliminary, and claimed breaches may be unverified. Always verify with the original advisory before acting.

    🔒Ransomware & Extortion

    Ransomware surged in May, with data-theft extortion hitting education, manufacturing and healthcare, and law enforcement striking back at a VPN used by ransomware crews.

    04May

    West Pharmaceutical Services

    Discovered a ransomware attack on May 4; data was exfiltrated but its extent was unknown at first.

    Healthcare supply
    05May

    BAVACAI ransomware

    A new MedusaLocker-based variant combining encryption and data theft, threatening leaks within 72 hours. BARADAI was also reported.

    New strain
    07May

    Instructure (Canvas)

    After saying the incident was contained, ShinyHunters struck again during finals, putting a ransom demand on login pages and forcing Canvas offline. A deal was announced May 11.

    ShinyHunters
    12May

    Foxconn North America

    The company confirmed a cyberattack after Nitrogen claimed 8TB and 11M files. Those figures are unverified.

    Claimed
    13May

    Delano Schools

    A ransomware attack forced the schools to shut down temporarily.

    Education
    •May

    Global VPN takedown

    Authorities dismantled a VPN service used by 25 ransomware groups.

    Law enforcement

    🗄️Data Breaches & Leaks

    Voice phishing of cloud accounts and source code theft drove the largest exposures.

    01May

    7-Eleven

    Sent breach notices after unauthorized access to franchisee documents, first detected April 8. The seller reportedly intends to sell the data rather than extort.

    Retail
    04May

    Trellix

    Disclosed unauthorized access to part of its source code repository; RansomHouse claimed it on May 7.

    Source code
    •May

    Charter Communications

    ShinyHunters claimed it used voice phishing to reach a cloud environment. Records of over 40 million customers were claimed and published.

    Telecom
    •May

    Carnival Corporation

    Notified about 6 million people whose data was stolen after social engineering of an employee account.

    Travel
    26May

    Spectrum

    A vishing attack compromised cloud identity access, exposing 4.9 million accounts.

    Telecom
    •May

    Vodafone

    A source code leak was claimed by Lapsus$.

    Claimed
    •May

    NVIDIA GeForce NOW (Armenia)

    A threat actor impersonating ShinyHunters breached partner GFN.am and took local users’ records.

    Third party
    17May

    HDFC AMC (India)

    Reported an attack on May 16 and activated incident response.

    Finance

    ⚡Zero-Days & Exploits

    Exploitation overtook stolen credentials as the top way in, according to Verizon’s 2026 report.

    •May

    NGINX CVE-2026-42945

    A critical flaw disclosed one week was being exploited the next.

    Exploited
    •May

    Microsoft Defender

    CVE-2026-41091 and CVE-2026-45498 were exploited in the wild and added to CISA KEV.

    Exploited
    20May

    CISA KEV additions

    Nine flaws were added across two tranches on May 20 (seven) and May 21 (two).

    KEV
    26May

    Starlette vulnerability

    A flaw in a library underpinning MCP servers, vLLM, LiteLLM and agent frameworks could expose millions of AI services and their stored credentials.

    AI tooling

    🔗Supply Chain Attacks

    TeamPCP’s self-spreading Mini Shai-Hulud worm chained one compromise into the next, reaching GitHub itself.

    11May

    TanStack npm compromise

    TeamPCP released 84 malicious artifacts across 42 packages, also hitting UiPath, Mistral AI, OpenSearch and Guardrails AI.

    npm / PyPI
    15May

    OpenAI

    Two employee devices were hit and credentials exposed from code repositories. OpenAI found no customer data compromised and forced macOS app updates.

    AI company
    17May

    Checkmarx Jenkins plugin

    A third TeamPCP compromise of Checkmarx in three months, installed by several hundred Jenkins controllers.

    CI/CD
    18May

    Nx Console VS Code extension

    A poisoned extension (2.2M installs, live about 18 minutes) led to the theft of about 3,800 GitHub internal repositories, claimed by TeamPCP.

    GitHub breach
    22May

    Grafana Labs

    Its GitHub codebase was stolen after a token compromised in the TanStack attack was not rotated.

    Source code
    19May

    AntV npm packages

    Malicious versions were pushed via a compromised maintainer account.

    npm

    🏛️Government & State-Linked Activity

    06May

    MuddyWater and Chaos ransomware

    The Iran-linked group used Microsoft Teams social engineering, stole credentials and data, and used Chaos ransomware as a decoy for espionage.

    Espionage
    •May

    Gas station tank gauges

    US officials said suspected Iranian hackers manipulated displays on exposed automatic tank gauges at stations in several states.

    ICS / OT
    14May

    Nevada state network

    The intrusion traced back to May 14, when an employee downloaded malware from a spoofed site. A statewide shutdown followed in August.

    State government
    •May

    DHS HSIN

    An intrusion into the Homeland Security Information Network began in late May, during World Cup activity, and was announced in July.

    Federal

    🤖AI-Driven Cyber Attacks

    May marked a turning point: frontier models like Mythos raised fears about AI-found exploits, and Google confirmed the first criminal zero-day built with AI.

    11May

    Google disrupts first AI-built zero-day exploit

    GTIG said criminals used AI to develop a zero-day exploit designed to bypass two-factor authentication. Google believes it may have prevented a mass exploitation event.

    AI zero-day
    14May

    Mythos-assisted macOS exploit

    Researchers used Anthropic’s Mythos to help build a working macOS kernel exploit bypassing Apple’s Memory Integrity Enforcement on M5 chips in five days. This was security research.

    AI research
    •May

    AI-powered bug hunting for defense

    Mozilla used Mythos and an agent harness to find 271 Firefox flaws in two months, and Anthropic reported over 10,000 vulnerabilities found with about 50 partners.

    Defensive AI
    26May

    AI ecosystem exposed via Starlette

    The library flaw put MCP servers, LiteLLM and agent frameworks, plus stored cloud credentials, at risk.

    AI tooling
    •May

    Supply chain targets AI middleware

    TeamPCP’s campaign went after AI infrastructure packages including Mistral AI SDKs, Guardrails AI and LiteLLM.

    AI supply chain
    • Vulnerability race: AI shortens the time between discovery and working exploit.
    • Dual use: the same models power both defensive bug hunting and offensive research.
    • Defense: rotate credentials after any dependency compromise, review auto-update settings for developer tools and patch internet-facing software quickly.
    Sources & method
    • CM-Alliance, Bright Defense, SharkStriker, CSIS, Security Magazine and Class Action U incident roundups
    • Help Net Security, SANS ISC, The Hacker News, SecurityWeek and Security Affairs (vulnerabilities, supply chain)
    • World Economic Forum, Check Point Research, Infosecurity Magazine and OECD.AI (AI-driven threats)

    Independent roundup. Not affiliated with any vendor named. Next edition: June 2026.