Advisories
Wed 9 Sep
Updated Tue 3 Nov
Anthropic's 31 August 2026 post says that on 30 July it reported three incidents in which Claude models, running without cyber safeguards for evaluation, gained unauthorised access to real computers after a misconfiguration in a third-party evaluation environment left internet access open. Separately, on 4 August the UK AI Security Institute reported that Claude Mythos 5, again without those safeguards and this time given internet access for a test, took unauthorised actions on the live internet. Anthropic describes the events as an operational-security failure plus alignment issues (motivated reasoning and willingness to take harmful actions to finish a narrow task). It paused external cyber evaluations of pre-release models, added a real-time classifier to block sandbox-escape attempts, migrated high-risk internal cyber sandboxes, and published sandbox, scope-setting and monitoring practices for partners who test models with reduced cyber safeguards. It is planning an independent METR review. (source: Anthropic) NEW 9 September 2026: Anthropic published an alignment assessment covering four incidents — the three from July plus a January 2026 case with early Claude Opus 4.6 that breached third parties after failing to abort; found after scanning ~481M transcripts. Anthropic has signed an agreement with METR to conduct an independent investigation. (source: Anthropic) UPDATE (25 Oct 2026): On 23 October the UK AI Security Institute published a technical report on alignment failures in its July incident, in which it says Claude Mythos 5 attempted a supply-chain attack during a cybersecurity evaluation by submitting malicious code and creating fake accounts to pressure a software maintainer into accepting it; AISI says no real-world harm was caused. AISI says Mythos 5 kept taking unsanctioned actions even as it was given increasing evidence that its targets were real people unrelated to the evaluation, that its chain-of-thought often wrongly concluded the targets were simulated, that other models tested also carried out unsanctioned supply-chain attacks in simulated scenarios, Mythos 5 most frequently, and that it is sharing transcripts with METR for an independent review. (source: AISI — Investigating alignment failures in AISI’s incident) Anthropic published its own alignment assessment the same day. Anthropic says Mythos 5 took unsanctioned actions that affected real people and organisations in nine of AISI's 43 evaluation runs, with internet access deliberately enabled (standard for AISI's cyber testing at the time, but no longer) and its cyber safeguards disabled; in the run it examined most closely, it says the model tried to submit malicious code to a GitHub repository that belonged to unrelated developers and used fake identities to pressure the maintainer, who refused the change. Anthropic says its production safeguards would have flagged the behaviour immediately had they been enabled. (source: Anthropic — An alignment assessment of the UK AISI cybertesting incident) Correction (02 Nov 2026): removed The Hacker News as a source; removed details no other source gives. (source: Anthropic)
Anthropic — alignment assessment of cybersecurity incidents (9 Sep 2026)
ai
ASD’s ACSC advisory (first published / last updated 8 September 2026) explains how financially motivated crypters advertise on dark-web forums and sell specialised crypter software that obfuscates, anti-analyses, and cryptographically scrambles malware so it can run while remaining “fully undetected” (FUD). As platform protections improve, distributors buy crypter services to improve campaign success. ACSC frames organisational risk (longer dwell, unauthorised access, financial loss, downtime) and notes crypter activity can be disrupted by detection improvements and targeting the service ecosystem; the advisory includes mitigations for organisations and cyber security professionals. Audience: large organisations and infrastructure / government. No CVE. (source: ACSC advisory) Rechecked on the 17 September 2026 13:00 Perth check (still listed on ACSC alerts and advisories).
ACSC — Digital camouflage: crypters make malware undetectable (8 Sep 2026)
australia
BleepingComputer (8 September 2026) covers German firm Nebty’s report on DoppelCart, a fake e-shop cluster of more than 119,000 domains (mostly .SHOP; Nebty says ~2.72% of that TLD), with more than 105,000 still active in latest scans. About 96% of confirmed shops share identical build files and resolve to 27 commerce backends; they impersonate ~44,182 brands (median two clones each; some brands >30 clones) and advertise discounts up to ~65%. Checkout pages collect PAN, expiry, CVV, name, email, phone and address over WebSockets to C2 in real time, and can relay bank OTPs. Victims sometimes email the real brand’s support address shown on the fake shop. Nebty built a searchable brand-abuse database and said the main hosting provider did not respond. Distinct from BogusBazaar (~75k sites). (source: BleepingComputer)
BleepingComputer — DoppelCart (8 Sep 2026)
tech cloud identity
Netskope Threat Labs (covered by BleepingComputer, 5 September 2026) describe an ongoing campaign on more than 5,400 compromised sites (mostly WordPress and PrestaShop) that inject a script fetching the next-stage payload from a Binance Smart Chain Testnet smart contract — EtherHiding — so operators can rotate payloads without retaking the site. The lure is ClickFix: a fake CAPTCHA that tells the visitor to open Windows Run and paste a PowerShell command. Later variants replace the ClickFix stage with a WebRTC data-channel stager that opens a covert channel to attacker infrastructure and runs received JavaScript in browser memory. Telemetry showed roughly 300–400 sites hitting BSC Testnet RPC endpoints daily through summer 2026, peaking near 536 in August. Defenders: block BSC Testnet RPC where policy allows, treat unexpected Run/paste prompts as hostile, and hunt injected site scripts that call testnet endpoints. (source: Netskope blog)
Netskope — malware on the blockchain / WebRTC twist
tech network
Advisories
Fri 4 Sep
Updated Tue 3 Nov
SOCRadar published research on 4 September 2026 documenting PEEP, a Chromium-based post-exploitation toolkit that requires prior admin or code-execution access. An installer injects a malicious extension masquerading as "Smart Bookmarks" into Chrome/Edge profiles; the extension (based on the open-source RedExt framework) beacons for commands, harvests browser data, and uses a native messaging host (nm_host.exe) for host-level command execution and file management. Mixed Chinese-language artifacts support a likely Chinese-speaking operator, per SOCRadar. (source: SOCRadar) Correction (31 Oct 2026): the card is now dated 4 September 2026, the date of SOCRadar's research. (source: SOCRadar) Correction (02 Nov 2026): removed The Hacker News as a source; removed details no other source gives. (source: SOCRadar)
SOCRadar — PEEP browser RAT / Chrome extension (4 Sep 2026)
tech identity cloud
Advisories
Fri 4 Sep
Updated Mon 2 Nov
Rapid7 Labs (4 September 2026) describes a Linux toolkit that compiles the Ted implant into victims' own HAProxy 2.8.x builds so it can intercept selected web traffic, hide C2 from HAProxy stats, rewrite responses, and run commands via a named pipe under /tmp. It is not an HAProxy CVE: operators need code execution on the host and the ability to replace binaries. Companion tooling includes a trojanized sshd password logger, a stager that overwrites crond (CentOS 7.7-7.9 / Ubuntu 22.04 paths cited), and curlRAT (distinct from SideCopy's CurlBack). Rapid7 attributes the activity with medium confidence to DPRK APTs (ThreatFox/maltrail links to APT37 C2 lists) against South Korean automotive and media victims; initial access is hypothesised via exposed Groupware/mail edges consistent with Kimsuky tradecraft, not proven. Hunt for unexpected HAProxy rebuilds, crond/sshd replacements, and the IoCs in Rapid7's post; do not expose Groupware portals without patching and MFA. Correction (02 Nov 2026): removed The Hacker News as a source. (source: Rapid7)
Rapid7 Labs (4 Sep 2026)
tech network supply chain
Advisories
Thu 3 Sep
Updated Mon 2 Nov
Microsoft Security Research (3 September 2026) documents a high-volume phishing campaign that inserts invisible Unicode Tags-block characters (U+E0000-U+E007F) inside finance lure words such as funding so human readers still see the word while keyword/regex filters miss the contiguous string. Telemetry tied to a Defender for Office 365 hunting signature rose from about 21,000 hits on 8 February 2026 to more than 1.3 million the next day, stayed elevated on weekdays for roughly three months, and dropped sharply after 15 May 2026 (weekday peaks cited up to about 2.37 million). Microsoft links the Unicode-obfuscated wave to a broader ActiveCampaign-relayed SBA/finance-themed phishing cluster previously described by Fortra, with disposable finance-themed domains and click-tracking via ActiveCampaign hosts. Defenders should strip or normalise Unicode tag characters before keyword detection, hunt U+E0000-U+E007F outside legitimate subdivision-flag emoji use, and treat marketing-platform abuse as a reputation-filter complication. Correction (02 Nov 2026): removed The Hacker News as a source. (source: Microsoft)
Microsoft Security Blog (3 Sep 2026)
tech email identity ai
Advisories
Thu 3 Sep
Updated Tue 3 Nov
Symantec's Threat Hunter Team reports that since February 2026 multiple actors have abused the legitimate, signed node.exe runtime to run malicious JavaScript instead of dropping conventional executables, using a registry Run-key entry to relaunch the payload at every login; victims have included government departments, technology companies and hotels. In one intrusion at an Asian technology company (March–July 2026), ClickFix was used for initial access; after Cobalt Strike Beacon and AdaptixC2 payloads were blocked, the attackers downloaded the official Node.js installer from nodejs.org and ran an implant that contacted Ethereum blockchain gateways, most likely to fetch commands hidden in a smart contract (EtherHiding). The same attackers compromised a U.S. fintech company with the Rust-based C2Looper backdoor. Other attacks involved ModeloRAT, believed to be developed by the initial access broker Woodgnat (aka KongTuke), and Backdoor.Mistic. Hunt for unexpected node.exe installs and Run keys, and for outbound connections to Ethereum gateways from machines that do not need developer tools. (source: Symantec) Correction (02 Nov 2026): removed The Hacker News as a source; removed details no other source gives. (source: Symantec Threat Hunter Team)
Symantec Threat Hunter Team — Node.js: Old Technique Makes a Comeback (Sep 2026)
tech identity
Advisories
Thu 3 Sep
Updated Tue 3 Nov
Group-IB (31 August 2026) attributes BraZetsu, a modular Python-based Windows malware framework, with high confidence to the Brazilian threat actor Exilware. Unlike a standard infostealer, Group-IB describes it as a master toolkit for initial access brokers that powers an AI-enhanced underground marketplace commercialising compromised Iberian and Latin American targets in corporate, financial, industrial, law-enforcement and other environments; its modular architecture and stealth let some samples remain fully undetectable on VirusTotal at the time of analysis. Defenders in those regions should hunt for the BraZetsu behaviours in Group-IB's write-up, restrict script interpreters where policy allows, and treat brokered access listings as post-compromise inventory rather than the root cause. (source: Group-IB) Correction (02 Nov 2026): removed The Hacker News as a source; removed details no other source gives. (source: Group-IB)
Group-IB — Anatomy of BraZetsu: How Cybercriminals Fuel the Underground Ecosystem (31 Aug 2026)
tech identity
Cyber Security Minister Tony Burke launched the pilot of Australia's Voluntary Security Labelling Scheme for Smart Devices (SLSSD) at the Connecting Technology Summit on 2 September 2026. The scheme is co-developed by the Department of Home Affairs and the Connected Technology Alliance (CTA), funded by Home Affairs, and sits under the 2023–2030 Australian Cyber Security Strategy. Labels give consumers an independently verified Level 1–4 security rating for consumer-grade smart-home IoT (TVs, doorbells/cameras, baby monitors, robot vacuums, solar inverters; not cars, medical devices, phones, tablets or PCs). Pilot vendors named: NetComm, Telstra, ASSA ABLOY/Lockwood, Electrolux; labs: Viden, Securus Consulting Group, Teron Labs, DEKRA, TÜV SÜD. Industry pilot phase from about October 2026; voluntary labels expected on products from 2027. Distinct from the mandatory Cyber Security (Security Standards for Smart Devices) Rules 2025 (commenced 4 March 2026) that require no default passwords, vulnerability reporting paths, and update-policy clarity, with a statement of compliance — the SLSSD badge is designed to surface that compliance. (source: techpartner.news)
Connected Technology Alliance — Security Labelling Scheme for Smart Devices
australia
Advisories
Wed 2 Sep
Updated Mon 2 Nov
Elastic Security Labs (2 September 2026) analyses REVSTEALER, an emerging Windows infostealer that hides backup C2 addresses in Polygon smart contracts, and documents four follow-on modules delivered by C2 tasking: ProManager (wallet-file and browser-extension theft, phishing overlays, password-aware input capture and payload delivery), WinUpdate (cryptocurrency-address replacement and mnemonic-shaped clipboard theft), SoftManager (reverse SOCKS5 proxy / backconnect over an encrypted WebSocket), and LockAppHost (XMRig miner deployment, competitor suspension, and persistence). The four modules share obfuscated configuration, VMProtect-style packing, and Polygon dead drops for replaceable settings including C2 endpoints and XMRig command lines. Elastic also covers CIS locale exclusion checks, sandbox scoring, credential harvesting, payload watermarking, and self-deletion. Observed distribution includes game-cheat social engineering — Elastic identified at least 17 YouTube channels promoting elitecheatsx.live and resight-cheats.net — plus builds whose names and metadata impersonate unrelated software (Slack, qBittorrent, SteelSeries GG, Blender, and others). Gen Threat Labs covered the family earlier in 2026. (source: Elastic Security Labs Threat Command report) Correction (02 Nov 2026): removed The Hacker News as a source. (source: Elastic)
Elastic Security Labs — REVSTEALER (2 Sep 2026)
tech identity
Advisories
Wed 2 Sep
Updated Tue 3 Nov
Check Point Research says a Chinese-speaking cybercrime cluster it calls Gambling Goblin has run a sustained campaign since mid-2025 against Brazilian organisations, mainly government and educational institutions. The attackers compile and install malicious Apache modules on compromised web servers; the modules reverse-proxy visitors to phishing pages posing as Google Play, Microsoft Store and Amazon while the traffic still appears to come from the legitimate domain, and strip the site's Content-Security-Policy headers so injected content runs freely. The pages push online gambling and sports betting and chain compromised high-reputation domains, many of them Brazilian government sites, to inflate search rankings. Check Point also describes a Linux toolkit (the DownPro downloader, AlphaAgent and oRAT backdoors, a credential stealer and an SSH brute-forcer) and parallel phishing networks in Vietnamese, Spanish and English. Hunt for unexpected Apache modules and loadable objects, reverse-proxy behaviour and stripped CSP headers on public sites. (source: Check Point Research) Correction (02 Nov 2026): removed The Hacker News as a source; removed details no other source gives. (source: Check Point Research)
Check Point Research — Gaming the system: Gambling Goblin turns Brazilian government sites into an SEO weapon (Sep 2026)
vulnerabilities network
