The Google Threat Intelligence Group has detected the first known instance of a threat actor using an AI-developed zero-day exploit in the wild. While the attackers planned a wide-scale strike, our proactive counter-discovery may have prevented that from happening. This finding is part of our new report on AI-powered threats.
Perhaps the sharpest article I've read in a while on the current (and future) state of cybersecurity.
"We've been shielded from exploits not only by soundly engineered countermeasures but also by a scarcity of elite attention" ~ @tqbf
Brace for impact.
https://t.co/aujwCzxYbV
the watchers: how openai, the US government, and persona have been secretly running an identity surveillance system since nov 2023.
https://t.co/Zz04WDF8Lz
researched by @vmfunc, @MDLcsgo, @DziurwaF
We became an admin in the Fédération Internationale de l'Automobile's driver categorisation system, which allowed us to access the PII and password hashes of any rated driver, including Max Verstappen. 🏎️
https://t.co/vdX7OegqmW
How the NSA (Equation Group) allegedly hacked into China's Polytechnical University 👀
I analysed intelligence reports from Chinese cyber firms (360, Pangu, CVERC) to aggregate TTPs attributed to Equation Group.
🔗https://t.co/2dQuwx0lxN
A huge percentage of the Internet sits behind us, including small businesses and emergency resources in Spain. We've always been happy and willing to work with rights holders in conjunction with judicial bodies to protect their content. We have a clear process that works around the world to do that. The strategy of blocking broadly through ISPs based on IPs is bonkers because so much content, including emergency services content, can be behind any IP. The collateral damage is vast and is hurting Spanish citizens from accessing critical resources. It's only a matter of time before a Spanish citizen can't access a life-saving emergency resource because the rights holder in a football match refuses to send a limited request to block one resource versus a broad request to block a whole swath of the Internet. When that unfortunately and inevitably happens and harms lives, I'm confident policy makers and courts in Spain and elsewhere will make the right policy decision. Until then, it'll be up to users to make politicians clear on the risk. I pray no one dies.
Un gran porcentaje de Internet está respaldado por nosotros, incluyendo pequeñas empresas y recursos de emergencia en España. Siempre hemos estado dispuestos a colaborar con los titulares de derechos, junto con los organismos judiciales, para proteger su contenido. Contamos con un proceso claro y eficaz a nivel mundial para lograrlo. La estrategia de bloqueo generalizado a través de los ISP basados en IP es absurda, ya que gran parte del contenido, incluido el de los servicios de emergencia, puede estar protegido por cualquier IP. Los daños colaterales son cuantiosos y están impidiendo que los ciudadanos españoles accedan a recursos críticos. Es solo cuestión de tiempo antes de que un ciudadano español no pueda acceder a un recurso de emergencia vital porque el titular de los derechos en un partido de fútbol se niega a enviar una solicitud limitada para bloquear un recurso en lugar de una solicitud amplia para bloquear toda una franja de Internet. Cuando esto, desafortunadamente e inevitablemente, ocurra y perjudique vidas, confío en que los legisladores y los tribunales, tanto en España como en otros países, tomarán la decisión política correcta. Hasta entonces, dependerá de los usuarios dejar claro a los políticos sobre el riesgo. Rezo para que nadie muera.
Amazon has announced they're phasing out their checkout-less grocery stores.
The "Just Walk Out" technology, which was labeled as automatic, was actually 1,000+ Indian employees monitoring you as you walked through the store.
https://t.co/uZlOZJgvC0
Me complace anunciar el lanzamiento de Cybersecurity AI (CAI), un framework open source de agentes autónomos diseñado para abordar escenarios y ejercicios de ciberseguridad.
Junto a este framework, publicamos el paper:
📄:“CAI, a bug bounty-ready Cybersecurity AI”,
🧬: https://t.co/5B9mo23Pwc
una contribución al avance del estado del arte abierto en agentes aplicados a seguridad informática.
pip3 install cai-framework
cai
📊 Resultados destacados de CAI relatados en el paper:
• 🧪 Top 30 en HackTheBox España en <1 semana
• 🤖 Top 1 en agentes de IA del CTF competitivo “AI vs Humans” de HackTheBox
• ⚔️ Máquinas medium y hard resueltas de forma completamente autónoma
• 🧭 Máquinas insane con un mínimo nivel de Human-in-the-Loop
• 🐞 Bug bounties reales completados exitosamente
📚 En el artículo analizamos las capacidades reales de modelos fundacionales (SOTA LLMs) aplicados al pentesting, CTFs competitivos y escenarios ofensivos realistas.
🔗 CAI es un framework multiagente, modular, agnóstico al modelo y que ofrece un SDK de agentes para ciberseguridad, con soporte integrado para:
DeepSeek-V3, Qwen 2.5, LLaMA 3, GPT (3.5/4), Claude (Opus/Sonnet), Gemini, O1/O3… y otros modelos abiertos y propietarios.
🧰 Además, Ofrece una CLI interactiva donde el usuario puede:
• Configurar 🛠️ agentes, herramientas ofensivas, targets y entornos
• Seleccionar modelos 🔄 y patrones agénticos
• Compatibilidad con servidores MCP (Especialmente util en Ghidra y BurpSuite)
• Iniciar, Interrumpir, reconfigurar y retomar el flujo de ejecución de los agentes
Todo bajo una filosofía híbrida: 🧑💻 usabilidad de frameworks clásicos de ciberseguridad + 💬 interfaces conversacionales adaptadas a arquitecturas multiagente, con la IA como núcleo operativo.
🔬 Esta investigación ha sido posible gracias a la colaboración interdisciplinar de investigadores y apasionados en ingeniería inversa, IA generativa, red teaming y automatización.
📌 En breve publicaremos PoCs, tutoriales técnicos.
🔬 Código disponible en: https://t.co/AyTmYvo3cA
Unete a nuestro discord:
https://t.co/T35HhM3fFF
Mención especial: @vmayoralv@francisco_oca
🧵 1/3
I usually need facial verification for OSINT investigations, and almost every free service forces you to create an account.
So, I started exploring open source tech and discovered DeepFace - but it had no UI.
Problem solved: https://t.co/dPZAOUdkB1
This is a guide on how to communicate securely, practicing good OPSEC and COMSEC. It is written for PGP but the same core principles apply for Signal. Or any communication that needs to be secure.
https://t.co/E2OjDXBDz2
The post is finally live! Join me for an in-depth analysis of CVE-2023-22098, which was discovered by the outstanding @theflow0 . We'll break down the vuln, explore some virtio-net internals, set up a debugging environment, and develop a reliable PoC to escape VirtualBox. Enjoy!
And we've got a winner 🏆
Congratulations to @caveeroo for the winning meme! The team will be in touch soon to award you your VIP+ Annual subscription.
👀 Stay tuned for the next #UniversityCTF24 challenge this Friday.