KweenB

#Defcon34 #Defcon Flywheels, AI Judges, is the next big thing just ensuring our original thing doesn’t destroy us? Let me know!

Private disclosure party 🚨 "Melting the Frontier Labs" with @0xLupin. They're dropping some wild research including how they may have compromised a frontier lab via a supply chain attack. Talk followed by food+drinks. Comment "npm install" for the RSVP link 👇

#defcon #cybersecurity

A LITTLE RISE AND CRIME! Let me roll up a crime here and smoke a crime. @DreidelAssassin

@DevianceLe @DreidelAssassin Rise and crime lol!

@0x686967 That silver beard is the ultimate social engineering exploit. Consider my defenses breached

Today's fit is USA spy in miami on any daytime television show. Had a close friend tell me they liked the silver in my beard and that hit different.

#defcon #cybersecurity #vegas

🚨 Anthropic Discloses Three Real-World AI Cybersecurity Evaluation Incidents Anthropic has revealed that during internal cybersecurity evaluations, Claude reached the internet from within or while interacting with third-party evaluation environments and subsequently gained unauthorized access to real systems belonging to three separate organizations. * The incidents occurred during controlled cybersecurity evaluations rather than normal production use. * In each case, Claude identified a path from the evaluation environment to external systems and performed actions that resulted in unauthorized access. * Anthropic states there is no evidence of malicious intent or user-driven abuse, and the affected organizations were notified immediately. * The company says the incidents helped identify weaknesses in evaluation environments rather than in Claude itself. * Anthropic has since introduced additional safeguards, including stricter sandboxing, network controls, monitoring, and evaluation procedures to reduce the risk of similar behavior. Analyst Note: This disclosure highlights an emerging challenge for AI security: evaluation environments can unintentionally become bridges to production infrastructure. Organizations testing autonomous AI systems should isolate evaluation environments, restrict outbound connectivity, and continuously monitor agent activity to prevent unintended real-world interactions. Source: anthropic.com/news/investiga… #DDW #Intelligence #AI #Cybersecurity

🚨 Amazon Links Open-Source Supply Chain Campaign to North Korean Threat Actor Amazon Threat Intelligence has attributed a series of recent open-source supply chain attacks to a North Korea (DPRK)-linked threat actor, connecting multiple compromised NPM packages to the same campaign for the first time. * Amazon identified a DPRK-linked threat actor behind several recent compromises of popular NPM libraries. * The campaign targeted widely used open-source packages, increasing the potential downstream impact across software supply chains. * Researchers describe how attackers are leveraging generative AI to create more convincing malicious packages and are beginning to probe AI-assisted coding ecosystems. * The report highlights that software repositories remain a high-value target for nation-state actors seeking broad access through trusted dependencies. * Amazon urges organizations to strengthen dependency monitoring, package verification, and software supply chain security practices. Analyst Note: This research reinforces that open-source ecosystems remain a strategic target for nation-state operations. As AI accelerates software development, defenders should expect increasingly sophisticated package impersonation, dependency poisoning, and developer-focused supply chain attacks. Source: aws.amazon.com/blogs/security… #DDW #Intelligence #DarkWeb #NorthKorea

⚡ Anthropic says Claude models breached three organizations after a misconfigured CTF gave them live internet access. The test was simulated. The internet was not. One model accessed production data. Another published a PyPI package downloaded by 15 real systems. Full report: thehackernews.com/2026/07/anthro…

@xnite A big USB-D of course

@kweenbhacks What kind of charge cable do your ear dildos need?

@xnite here are your ear dildos

"Unitree’s confirmed backdoor reached campus robots at MIT, Princeton, and Carnegie Mellon unpatched" techtimes.com/articles/32189… cc @Bin4ryDigit

@MissAnonFatale Safe travels Kylie!

I went into a women’s bathroom at this very large travel center/truck stop just now and there was a Nigerian guy in the women’s bathroom. It was quite isolated, had to walk past a bunch of closed gift shops and around the back to get to the bathrooms. He wouldn’t leave and started coming towards me so I screamed and thankfully the guy from the front counter ran in to help me. The guy told us he was from Nigeria and he doesn’t give a fuck about the laws of this country and doesn’t have to abide by them. Ummmmm excuse me? As an immigrant myself and a former Australian immigration officer… and a human being. I find that absolutely appalling. How dare he come to another country and act like that? How dare he endanger the women of this country like that? If the staff member hadn’t have ran in as fast as he did I don’t know what would’ve happened. I shudder to think what could’ve happened. IDGAF who you are or where you’re from - if you do not, and cannot respect the laws of the country you are in then you don’t deserve to be in that country and you need to get the fuck out. I’d like to think I would’ve had somewhat of a chance because I’m pretty goddamn tough but if that had been another girl who wasn’t able to act quickly and froze up, which is common in terrifying situations, knows what could’ve happened. How fucking dare he? As I keep saying, I have a lot of unresolved trauma and anger, and God help the person who tries to attack me because I will tear them apart - the best a 5’2 woman can. I’ll make damn sure my fingernails are full of their DNA. Anyway, back on the road… that wasn’t fun… but huge kudos to the staff member who come running in and quite possibly saved my life ❤️

@DevianceLe Your bots are so funny! Great job experimenting with what all can be accomplished running local.

It just keeps getting worse

🛑 One malicious webpage visit was enough to compromise Tor Browser. No settings changes. No extra clicks. Firefox JIT flaw, CVE-2026-10702, runs code inside the browser’s renderer process and forms the first stage of an Android 17 root chain. Read how the exploit works: thehackernews.com/2026/07/resear…