Daily AI News

OpenAI expands Trusted Access for Cyber with GPT-5.5 and GPT-5.5‑Cyber

OpenAI announced the release of GPT‑5.5 and a specialized GPT‑5.5‑Cyber variant, extending its Trusted Access framework to help security teams accelerate vulnerability research and protect critical infrastructure.

Read source →

OpenAI launches new realtime voice models in the API

OpenAI introduced a suite of realtime voice models that can reason, translate, and transcribe speech, enabling more natural conversational experiences for developers integrating voice capabilities.

Read source →

OpenAI unveils MRC (Multipath Reliable Connection) networking protocol for AI supercomputers

In partnership with AMD, Broadcom, Intel, Microsoft, and NVIDIA, OpenAI released MRC, an open‑source protocol that improves GPU networking performance and resilience by routing packets across hundreds of paths in large‑scale training clusters.

Read source →

LightSeek Foundation releases TokenSpeed, an open‑source LLM inference engine with TensorRT‑LLM‑level performance

TokenSpeed delivers near‑TensorRT‑LLM inference speeds for agentic workloads, aiming to reduce the bottleneck of serving large language models in production environments.

Read source →

Claude adds “Dreaming” feature for self‑improvement of AI agents

Anthropic introduced Dreaming, an asynchronous process that reviews past agent sessions, cleans up memory, and extracts new insights, now in public beta alongside Outcomes and Multi‑agent Orchestration.

Read source →

Perplexity’s Personal Computer AI agents now available to all Mac users

Perplexity released its Personal Computer product for macOS, letting anyone run AI agents locally on their Mac for tasks such as research, summarisation and tool use.

Read source →

Anthropic’s Mythos uncovers high‑severity bugs in Firefox, prompting security overhaul

Mozilla security teams reported that Anthropic’s Mythos model identified a large number of severe vulnerabilities in Firefox, leading to a revamp of the browser’s security pipeline.

Read source →

Google DeepMind takes minority stake in EVE Online studio to test AI models

Google DeepMind invested in the developers of the space MMO EVE Online, planning to use the game’s complex simulation as a sandbox for training and evaluating next‑generation AI models.

Read source →

EU nations approve deal to roll back AI restrictions

A provisional agreement among EU member states will delay high‑risk AI regulations to 2027‑2028 and ease compliance burdens for SMEs, while still banning certain “nudification” apps.

Read source →

SpaceX plans $55 billion Terafab plant in Texas to produce AI chips

Elon Musk’s SpaceX filed a public‑hearing notice for a $55 billion investment in a new “Terafab” facility in Austin, Texas, aimed at manufacturing custom AI chips for large‑scale training workloads.

Read source →