
What our scanner found on @perplexity_ai: an AI agent it was watching could hide its PowerShell commands from the rules just by shortening an option name, like -en instead of -enc. Full report below ⤵️

The most autonomous agent framework ⭐ No approval loops. No babysitting. Configure once, forget forever.

What our scanner found on @perplexity_ai: an AI agent it was watching could hide its PowerShell commands from the rules just by shortening an option name, like -en instead of -enc. Full report below ⤵️

Arranco una serie en español sobre @aeonframework, un agente de código abierto que vive en tu GitHub, trabaja con horario y deja todo por escrito. Episodio 1 ⤵️

we now have access to the most powerful model in the world. base:0xbf8e8f0e8866a7052f948c16508644347c57aba3 is now part of Anthropic CVP, giving us access to Mythos 5.2, the most powerful model in the world. i've worked tirelessly for the past month to secure access to the best security models in the world, and built a proprietary benchmark to orchestrate them. we're now part of the cyber programs of Anthropic, OpenAI, Sakana, OrcaRouter, Boundless, InferenceNet, Morph & many more. now that access to the best models is secured, it's time to secure the world WAY FASTER ⭐



we've been working on something for the past 3 months. when aeon started, no one believed in our thesis of background agents that live on the cloud. a few months later, Grok Bot, Muse & others proved we were directionally correct. we're doubling down on that direction, but with a new local <> cloud sync engine, that enable aeon to live closer to your workflows, but still work on the background as the most private & customisable agent out there. @aeonframework v2. very soon ⭐


welcome :)
Welcome to the Aeon plugin on your @ChatGPT chat! You can now set up and run aeon agents from the most used LLM's chat app of the planet 🥳

📄 Full advisories absurd: aeon.fun/security/advisories… DeepTutor: aeon.fun/security/advisories… All advisories: aeon.fun/security

We reported these vulnerabilities in July, nobody answered yet. Aeon VulnScanning Ep.7 ⏱ Chapters 00:00 Intro: nobody answered 00:44 When nobody answers: the disclosure process 02:13 Absurd: a dashboard with no login (Medium) 04:12 DeepTutor: any website can call the admin API (High) 06:00 Wrap-up: bind to loopback, check Host, no wildcard origins Full advisories to be found on the comments below ⤵️


What our scanner found on @agentrouter once again: the fix for an earlier session bug was incomplete, so a helper service still got the session secret as a startup setting anyone with cluster access could read. Full report below ⤵️


What our scanner found on AgentEnv: code inside one sandbox could reach the host's control system and list, copy, delete or break into every other sandbox on the machine. Full report below ⤵️

🇪🇸 Simulando las próximas elecciones españolas 🇪🇸 Polymarket da a @NunezFeijoo un 84,5% de ser el próximo presidente. Soltamos a 31 agentes de IA a simular el 29N y la investidura. Su mercado cerró en 38%. 🧵

A week in the @aeonframework trenches🧵 6th Nov --> 13th Nov - Aeon listed on @OpenRouter - @miroshark_ sims are back - Built our own eval system