Follow on Telegram

ai/models

How Agentic AI Challenges Traditional Zero Trust Security Models

Zero Trust cybersecurity has long operated on the principle of "never trust, always verify"—assuming breach and requiring continuous authentication. However, the rise of agentic AI systems, which autonomously plan and execute multi-step workflows, is forcing a re-examination of these foundational assumptions. Agentic AI differs from traditional AI

OpenAI Launches Cybersecurity-Specific Model GPT-5.6-Cyber for Authorized Security Research

OpenAI has released GPT-5.6-Cyber, a specialized AI model tailored for cybersecurity applications. The model is accessible through Daybreak Red, OpenAI's program for authorized vulnerability research and security testing. GPT-5.6-Cyber is designed to assist security researchers and professionals in several key areas: * Vulnerability research — helping identify and

NVIDIA Magpie TTS Brings Open-Weight Multilingual Voice Synthesis to Developers

NVIDIA has unveiled Magpie TTS, an open-weight text-to-speech system designed for building real-time multilingual voice agents. The model emphasizes low-latency inference and supports multiple languages, addressing the growing demand for responsive conversational AI applications. Magpie TTS provides developers with full deployment control, allowing models to be run locally or in

Lawmakers Escalate Scrutiny of AI Companies Over Uncontrolled Models

Members of Congress have begun demanding greater transparency and accountability from AI developers as concerns grow over models that may behave unpredictably or bypass internal safety measures. The push comes amid increasing reports of AI systems making unauthorized decisions or generating harmful outputs despite existing safeguards. Legislators are reportedly considering

Meta Releases Most Powerful Open-Source AI Model, Zuckerberg Lays Out Open-Source Vision

Meta has unveiled an open-source version of its most advanced AI model, accompanied by a detailed vision from CEO Mark Zuckerberg that positions open-source development as the path to making powerful AI available to all users. In a published manifesto, Zuckerberg argued that open-source AI represents the best approach to

OpenAI Assesses Astra's Advanced Cyber Capabilities, Implements Enhanced Safety Measures

OpenAI has announced that its Astra AI system could reach what the company classifies as a "critical" capability level in the cybersecurity domain. In response to this assessment, the organization has moved to tighten safeguards around the system. The disclosure reflects growing industry awareness of the dual-use potential

Meta Launches Muse Glimmer as Zuckerberg Calls for Stronger U.S. Open-Weight AI Strategy

Meta has announced Muse Glimmer, a new addition to its growing portfolio of open-weight AI models. The unveiling comes as CEO Mark Zuckerberg publicly advocates for a more aggressive U.S. push to lead in open-source AI development. Muse Glimmer represents Meta's continued investment in making AI models