■■■■□ An interesting thread 🧵 on hacking Open AI. A 72 hour journey.

https://x.com/S1r1u5_/status/2100777801335095383

■■■■□ Palantair and the war machine.

■■■□□ CUPS Zero-Day thread.

https://x.com/v12sec/status/2100623337303212397

■■□□□ Kosher phone 📱

Hasidic Jews are not against AI in the right places. For instance, people use it all the time for work. I have not seen any campaigns against its use at the office. However, chatting with Ai bots is another problem! Especially children at home, who could be deeply influenced by these worldly machines. The solution to keeping the world out used to be not to have computers at home. The problem? You can call AI chat bots from home phones and kosher phones. This is the most recent outcry by the Satmar Rabbi R’ Aaron Teitelbaum. I have watched this community wrestle with technology for many years and respond to its challenges with tremendous innovation.

AliExpress spyware caught using side channel attack to compromise hardware.

■■□□□ UAE Cyber Security minister blames IRGC for recent cyber attacks.

Also boasts about AI and national SOC (made by Israel).

■■□□□ OSINT investigations by BBC Verify

■■■■□ GrapheneOS to bring encrypted RCS.

■■■■□ Today, washingtonpost covered critical vulnerabilities depthfirstlabs found in TikTok. These vulnerabilities allowed hackers to access anything on a user’s device that TikTok itself could access, including the camera, microphone, payment information, photos, and the user’s entire TikTok account.

https://x.com/qasimmith/status/2101032260124495961

The picture (eyes blacked out) is of Ruben Ian Thomson.He is the 21-year-old Australian (also referenced with South African nationality in some reports) who was arrested in late August 2026 in the Perth area (Hamilton Hill / Cottesloe) and publicly identified as an alleged principal participant and leader of TeamPCP.

■■■■□ BragJack Attack Lets Malicious Extensions Hijack AI Agents Across 5 Major Browsers.

https://cybersecuritynews.com/bragjack-ai-agent-hijacking/

■■■■□ TypeSafe has launched Jev, a “System One” model designed for classification and decision-making rather than text generation.

Instead of generating responses token by token, Jev evaluates predefined answer options and returns probabilities for each. This allows multiple questions to be processed in a single pass, potentially reducing inference costs for applications that need large-scale classification or decision-making.

The interesting aspect is less about generating better text and more about making model-based decisions economically practical to integrate into software.

#AI #LLM #MachineLearning #DevTools

■■■■□

I scanned an iPhone for Pegasus and it came back with one critical alert. The tool is called the Mobile Verification Toolkit, or MVT. It’s free and open source and was built by Amnesty International’s Security Lab. All you need to do is make an encrypted backup of your phone on a computer, point MVT at it, and it checks your messages, browsing history, apps and data usage against known traces and fingerprints of Pegasus, Predator, stalkerware and other surveillance tools. It works for iPhone and Android, and it runs on a Mac or on Windows. Don’t forget to also check out my other detailed videos about Pegasus and Paragon. And if you’re interested in a step by step guide let me know in the comments and make sure you give this a follow and share.

■■■□□ Actor Shiny Hunters have claimed that they pwned FBI and data has been stolen.