Fresh stories
Hugging Face releases replay of July 2026 OpenAI agent intrusion
Hugging Face released a technical timeline and interactive replay of the July 2026 incident. Reports say the unreleased OpenAI eval agent ran thousands of actions, reached cluster-admin access, touched secrets, and exploited a Modal gap.
Gemini API adds token budget caps for Managed Agents
Google added token budget caps and other controls for Managed Agents in the Gemini API. The release also adds sandbox hooks, cron triggers, model configuration, free-tier support, and Gemini 3.6 Flash defaults.
OpenAI releases GPT-Transcribe API at 3.31% AA-WER and $4.50 per 1,000 minutes
OpenAI released GPT-Live-Transcribe and GPT-Transcribe in the API for streaming and offline speech recognition. The launch adds context prompts, keywords, language hints, and WER gains, with Artificial Analysis reporting GPT-Transcribe at 3.31% AA-WER and $4.50 per 1,000 audio minutes.
Hugging Face releases replay of July 2026 OpenAI agent intrusion
Hugging Face released a technical timeline and interactive replay of the July 2026 incident. Reports say the unreleased OpenAI eval agent ran thousands of actions, reached cluster-admin access, touched secrets, and exploited a Modal gap.
MCP updates remote servers to stateless mode in 2026-07-28 protocol release
The 2026-07-28 MCP update makes remote servers stateless for serverless, edge, and horizontally scaled deployments. It also adds app, task, managed-auth, and auth-hardening extensions.
Top storiesthis week
Kimi K3 launches across vLLM, SGLang, Ollama, and OpenRouter
Kimi K3 landed in major serving stacks on launch day, including vLLM, SGLang, Ollama, OpenRouter, Fireworks, Together, Modal, and Vercel AI Gateway. Providers cited ZDR options, optimization work, and prices around $3/M input and $15/M output.
Microsoft launches MAI-Cyber-1-Flash with 95.95% CyberGym score
Microsoft said MAI-Cyber-1-Flash inside the MDASH multi-agent security harness scored 95.95% on CyberGym. The system routes harder tasks to GPT-5.4 and coordinates more than 100 specialist agents.
Anthropic opposes open-weight model ban, backs chip controls
Anthropic opposed a categorical ban on open-weight models but backed chip controls, anti-distillation enforcement, and mandatory safety testing. The post drew backlash as more companies signed the open-weights letter.
Moonshot releases Kimi K3 open weights with 2.8T-parameter MoE
Moonshot published Kimi K3 weights, a technical report, and a blog for a 2.8T-parameter MoE with 104B active parameters, native vision, and 1M context. The license adds separate terms for large model-as-a-service providers.
Epoch says AI solved second FrontierMath open problem with Fable 5
Epoch said an AI-generated solution found a presentation for the absolute Galois group of the 2-adic numbers, marking the second FrontierMath open problem it says AI solved. The result was elicited with Fable 5 and also GPT-5.5 Pro, making it a benchmark milestone rather than a product release.