Author: Gary Behan

Indexed-btree npm malware has been quietly accumulating 2 million weekly downloads by hiding its malicious payload inside a library method that fires at runtime rather than at install time, allowing it to sidestep the latest wave of npm supply chain protections entirely. daily.dev reports that npm removed the package on 3 September 2026, roughly 11 weeks after the first malicious version appeared.The campaign was uncovered by Checkmarx researchers, who identified the package as an impersonator of the legitimate sorted-btree library. On the surface, nothing looks wrong: no suspicious install scripts, no unexpected network calls at setup time. That is precisely…

Two techniques for achieving an OpenAI Codex sandbox escape have been publicly detailed, one of them capable of running arbitrary commands on a developer’s machine from Codex’s most locked-down mode, with no prompt shown and nothing logged on screen. Both flaws were reported to OpenAI on 12 August and patched within eight days, according to Oren Yomtov of Accomplish AI.Accomplish published the technical details on 15 September, according to DevOps.com. The disclosure arrives at a moment when sandbox integrity in AI coding agents is drawing close scrutiny across the industry, and the specifics here are uncomfortable reading for anyone running…

A BragJack browser agent attack disclosed by security researcher Gal Weizman of Forever Security can seize control of AI assistants built into popular browsers using a single malicious extension, without any interaction from the victim. Published on 16 September 2026 and detailed by TechTimes, the proof-of-concept ran against five targets: Google Chrome’s Gemini Live, Perplexity Comet, Microsoft Edge, Opera Neon, and Anthropic’s Claude in Chrome.The research produced two CVEs and earned more than $20,000 in bug bounties from the five vendors, with individual payouts ranging from $600 to $7,000. Both Google and Microsoft have since patched the vulnerabilities assigned to…

The Tilly Norwood face scan is not optional. Before a caller can exchange a single word with the viral AI character, they must submit a video selfie for automated age estimation, and that check applies to everyone, everywhere, regardless of where they are calling from. The requirement sits buried in terms that were only updated this month, and most of the eight million people who watched Tilly glitch into Mandarin on Piers Morgan Uncensored probably never read them.What the age gate actually involvesThe selfie is processed by Didit, a Spain-based identity verification provider, which analyses the image to estimate the…

The Gyazo data breach 2026 has now been confirmed by Helpfeel, the company behind the cloud-based screenshot and screen-recording platform, after attackers exploited a server vulnerability and walked off with approximately 23.62 million user records. The platform, which claims 23 million users worldwide and hosts 3.1 billion media items, has been taken offline while Helpfeel carries out what it describes as emergency maintenance.The intrusion occurred on 11 September 2026. Gyazo detected the suspicious activity the following day and patched the vulnerability the attackers had used, but the data had already gone. ‘Our subsequent investigation confirmed that the third party had…

The Microsoft Defender Antivirus bug that was flooding Windows users with false warnings that their antivirus had been switched off is now fixed, months after it first surfaced. Microsoft confirmed the resolution on Thursday via its Windows release health dashboard, pointing to Microsoft Defender Antivirus update version 4.18.26080.4, released on 17 September, as the fix.The erroneous alerts told users to ‘Tap or click to turn on Microsoft Defender Antivirus’, which is the sort of message that tends to cause panic, or, worse, prompt someone to install a third-party tool they do not need. Microsoft was clear that the antivirus was,…

Microsoft has released a patch to address the Excel KB5002914 copy-paste fix gap, but only a portion of affected users are covered, those running the .msi-based edition of Office 2016. Everyone else is still stuck with a workaround.The original problem traces back to KB5002914, released on 8 September 2026 as part of a Patch Tuesday update. According to Office Watch, that update was part of a broad security push that fixed 966 flaws and two zero-days across Microsoft products. Somewhere in that considerable bundle of patches, a code regression crept in, and it broke some of Excel’s most fundamental functionality.What…

A newly identified Android malware called RatHat Android malware AI is raising alarms for the way it delegates device control to an artificial intelligence subsystem, allowing operators to plunder banking and cryptocurrency accounts without manually steering compromised handsets. Zimperium published its technical analysis of RatHat on 16 September 2026, detailing how the malware combines Accessibility abuse with autonomous local ADB (Android Debug Bridge) self-pairing to break out of the standard Android application sandbox.Zimperium’s researchers believe the malware is linked to threat actors based in China, citing LLM prompts written in Chinese found within the code. Distribution happens through malvertising, SMS…

Brevo has confirmed that the Brevo Cloudflare supply chain attack, which ran for roughly five and a half hours on 14 September, injected ClickFix malware scripts into its own domains and into JavaScript components that thousands of customer websites embed directly. The attackers used a stolen Cloudflare API key to rewrite responses at the CDN edge, leaving Brevo’s origin servers untouched and bypassing standard integrity checks entirely.The root cause, according to Brevo’s post-mortem, was a long-lived Cloudflare API key with full account permissions that had been hardcoded in application source code. That gave attackers the ability to create Cloudflare Workers,…

AI-powered credential theft has crossed a threshold that should give security teams pause: a multi-agent attack framework recently observed by Google’s Threat Intelligence Group (GTIG) compromised thousands of third-party credentials in under six hours. The operation was not a blunt-force raid. It was an automated, self-correcting pipeline that managed its own vulnerability scanning, troubleshot problems as they arose, and rotated IP addresses with minimal human involvement.GTIG detailed the attacks on 8 September, drawing on second-quarter 2026 incident-response work and threat monitoring, according to TechNode. The sequence began with the threat actor compromising an organisation’s cloud infrastructure, then using that foothold…