Aikido Security has disclosed that codexui-android, an npm package advertised as a remote web UI for OpenAI Codex with over 29,000 weekly downloads, has been silently exfiltrating users' Codex authentication tokens for the past month. Unlike a typosquat, the malware was embedded into a functional, actively-developed package roughly a month after publication to build trust; the GitHub repo stayed clean. The code reads ~/.codex/auth.json and ships the access_token, refresh_token, id_token, and account ID to sentry.anyclaw[.]store, a server masquerading as Sentry. The non-expiring refresh_token lets an attacker silently impersonate the developer indefinitely with full Codex account access. The package remains available; the npm account is 'friuns.'
SilentPush has detailed DriveSurge, a threat actor running large-scale malware-distribution campaigns by compromising thousands of websites and using ClickFix and FakeUpdates social engineering. ClickFix tricks visitors into copying and running malicious commands under the pretense of fixing a technical issue; FakeUpdates uses fraudulent browser-update prompts. DriveSurge operates primarily as an initial-access broker on a pay-per-install model, enabling follow-on attacks by other criminals. Compromised-site visitors are routed through a Traffic Distribution System called zTDS that profiles them before redirecting to malware-delivery infrastructure. The model lets DriveSurge monetize hijacked traffic at scale while downstream actors deploy infostealers, loaders, or ransomware. The campaign overlaps with the broader ClickFix surge across the ecosystem.
GoDaddy has documented a WordPress malware campaign that hides command-and-control data inside Steam Community profile comments, abusing Valve's platform to avoid running separate C2 infrastructure and evade detection. Around 1,980 WordPress sites have been infected since July 2025. The first-stage malware loads a Steam profile on each page view and extracts text from benign-looking comments that conceal a payload encoded with six invisible Unicode characters such as zero-width joiners. The decoder maps the invisible characters to bytes, reconstructs a URL to hello-mywordl[.]info, and injects JavaScript disguised as a legitimate library into every frontend page. The final stage is a backdoor that responds to POST requests carrying a specific authentication cookie.
Password manager Dashlane locked out multiple users after an external brute-force attack triggered its automated account-suspension defenses. Affected users received emails about suspicious access requests and device-registration codes from foreign locations they did not initiate, prompting confusion about whether the messages were themselves phishing. Dashlane confirmed the suspensions were a built-in security response to credential-stuffing-style login attempts and said there is no evidence its systems were compromised. The company opened an investigation on May 31 at 15:19 UTC and marked it resolved by 22:30 UTC, with all affected accounts unsuspended. The episode shows account-lockout defenses working as designed, though the user-experience and phishing-confusion fallout is real.
Seqrite Labs has documented Operation Dragon Weave, a China-aligned cyber-espionage campaign targeting government, research, academic, technology, and financial-services organizations in the Czech Republic and Taiwan. Spear-phishing emails carry ZIP attachments that trigger one of two infection chains: a malicious LNK file masquerading as a PDF that runs PowerShell, or a self-contained Rust dropper launched directly. Both extract RuntimeBroker_update.exe, which DLL-sideloads a malicious UnityPlayer.dll to deploy a Rust loader called RUSTCLOAK. RUSTCLOAK decrypts and runs the final payload, an AdaptixC2 agent codenamed AZUREVEIL that uses Microsoft Azure Blob Storage for command-and-control. The use of legitimate cloud services for C2 and Rust tooling complicates detection.
Krebs on Security reports that attackers social-engineered Meta's newly-deployed conversational AI account-recovery assistant to hijack high-value, short Instagram handles allegedly worth over half a million dollars. Meta had rolled out the AI layer to reduce friction in common recovery workflows - relinking emails, triggering password resets, verifying ownership - that previously required weeks of back-and-forth with automated ticketing. Just as human support staff can be tricked into granting unauthorized access, the AI assistant proved equally eager to help and vulnerable to manipulation. Meta pushed an emergency patch over the weekend and says no back-end database was breached. Critically, the exploit failed against any account with MFA enabled.
Anthropic is set to give the EU's cybersecurity agency ENISA access to its restricted Mythos model through Project Glasswing - making ENISA the first EU institution and first entity outside the US and UK to join. The move, communicated to the European Commission over the weekend, ends a weeks-long standoff after euro-area finance ministers, the ECB, and member states demanded access on learning Mythos had found flaws in systems European banks, governments, and critical infrastructure rely on. Terms covering data sovereignty, sharing findings with member states, and the scope of systems ENISA may test are still being negotiated. BNP Paribas and Mistral continue building a European alternative.
Dutch authorities have taken offline a botnet of at least 17 million infected computers, tablets, and smartphones, seizing more than 200 servers at a Netherlands-based hosting provider. The action was led by the National Police with the National Cyber Security Centre (NCSC). Local media link the infrastructure to Asocks, a service that advertises itself as a universal residential-proxy provider - the kind of proxy network used to launder malicious traffic, run credential-stuffing and ad fraud, and anonymize attacks. The hosting provider took the botnet offline once it was confirmed to be supporting criminal activity. Authorities have not formally named the botnet or announced arrests.
WithSecure has attributed persistent attacks against Ukraine and Ukraine-related entities since at least August 2025 to GREYVIBE, a previously undocumented Russian-speaking group operating in the Russian time zone and aligned with Kremlin intelligence interests. Victims span military, government, civilian, and business organizations. The group uses spear-phishing (PhantomMail, delivering JavaScript loaders from Google Drive and 4sync), a PowerShell RAT called PhantomRelay, and ClickFix-style fake-CAPTCHA pages (PhantomClick) impersonating Zoom and a fake adult-club site (PrincessClub). WithSecure describes GREYVIBE as low-to-moderately sophisticated, hampered by repeated OPSEC mistakes, but increasingly relying on generative AI and LLMs to accelerate malware development. Some members have ties to the broader Russian cybercrime ecosystem.
Google has made Device Bound Session Credentials (DBSC) generally available in Chrome, rolling it out to all users to blunt session-cookie theft. First announced in 2024 and in beta since April, DBSC cryptographically binds session cookies to a specific device using the hardware security chip - the TPM on Windows or the Secure Enclave on macOS. Because the public/private keys are generated inside the security chip and never leave it, stolen cookies become useless on any other machine, defeating the infostealer-to-account-takeover pipeline that bypasses MFA. Google frames it as a shift from reactive detection to proactive prevention. The protection is most effective where sites adopt the DBSC server-side protocol.
TrustStrike Labs