Wiredframe Radar logo

Category intelligence

AI News Briefing — August 13, 2026

95 current items analyzed and ranked.

Executive synthesis

AI News Summary

Executive Signal

  • Frontier-class capability, regulatory scope, and prompt-IP exposure have all shifted within one news cycle — enterprises must revisit the open-vs-closed build decision before the next planning horizon.

Priority Developments

  • Open-weights reach Max parity: Qwen3.8-2.4T-A95B released as open weights with Day-0 vLLM and FP8/MXFP4 support; collapses the capability argument against self-hosting frontier models.
  • Regulation extends to open models: White House framework reportedly brings open models under regulatory scope, eroding the compliance advantage that historically favored open deployment.
  • Prompt confidentiality is broken: IIT Bombay/Adobe inversion model reconstructs system prompts from outputs at near-perfect accuracy — IP embedded in prompts is no longer defensible.
  • Capital concentrates in the AI application layer: Thrive Holdings ($2B at $12B, OpenAI-backed) and Lovable ($400M at $13.3B) signal investor conviction that value capture is shifting up the stack from foundation models.

Leadership Implications

  • Re-run the open-vs-closed model sourcing decision within 90 days; capability, compliance, and IP exposure have moved simultaneously.
  • Migrate proprietary prompt IP into weights, fine-tunes, or encrypted inference paths before prompt-inversion becomes standard adversary tooling.

Key Themes

Open-source model releases · 5Frontier Model Releases and Benchmark Competition · 3Model Releases · 3AI Policy & Regulation · 8AI Security and Interpretability Research · 2AI Security & Safety · 5Open Source / Open Weights · 3AI Funding & Startups · 5Market Share and Competitive Dynamics · 2Healthcare & Research AI · 1

Primary evidence

Top Ranked Signals

News huggingface.co 5 days ago

Qwen/Qwen3.8-2.4T-A95B · Hugging Face

90 score
AI Analysis

Qwen released Qwen3.8-2.4T-A95B on Hugging Face, a 2.4-trillion-parameter MoE model with 95B active parameters, available in Transformers format compatible with vLLM, SGLang, and TokenSpeed. It is the open-weights counterpart to Qwen3.8-Max and is positioned as the most capable generation in the Qwen open family.

Qwen3.8-2.4T-A95B This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format. These artifacts are compatible with vLLM, SGLang, TokenSpeed, etc. For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided by Qwen Cloud . In particular, Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B with more features, such as vision input & non-thinking support, 1
open_source_releaseqwenmoefrontier_models
News vllm.ai 5 days ago

Day 0 Support for Qwen3.8-2.4T-A95B on vLLM | vLLM Blog

88 score
AI Analysis

vLLM announces Day-0 support for Qwen3.8-2.4T-A95B, the first Qwen-Max-class model released as open weights, with FP8 and BF16 official checkpoints plus Inferact MXFP4/NVFP4 quantized variants. Inference requires at least two NVIDIA B3 GPUs.

Table of Contents We are announcing Day-0 vLLM support for Qwen3.8-2.4T-A95B. This is the first model from the Qwen family to bring a Qwen-Max-class model to open-weight release. Qwen3.8-2.4T-A95B is built on the Qwen 3.5 architecture and runs on vLLM out of the box. In addition to the official FP8 and BF16 checkpoints, Inferact has released MXFP4 and NVFP4-quantized weights that match full-precision quality while significantly reducing memory and bandwidth overhead. Qwen3.8-2.4T-A95B is a 2.4-t
open-sourcemodel-releaseinferencemoeqwen
News Feed: Artificial Intelligence Latest 6 days ago

The White House Is Going to Expand Its AI Policy

By Hugo Lowell

78 score
AI Analysis

The White House is preparing an updated AI framework that, according to sources, will bring open models under regulatory scope. The move reflects continued efforts to shape AI policy without formal legislation.

Open models may soon be added to an updated AI framework, sources tell WIRED, as the White House continues to grapple with how to regulate a technology it has tried not to regulate.
ai_policyus_governmentopen_sourceregulation
News AI News & Artificial Intelligence | TechCrunch 6 days ago

OpenAI-backed Thrive Holdings raises $2B to bring AI to the enterprise

By Rebecca Bellan

73 score
AI Analysis

OpenAI-backed Thrive Holdings has raised $2 billion at a $12 billion valuation from SoftBank, D1 Capital Partners, and Altimeter Capital to bring AI to the enterprise.

Thrive Holdings has raised $2 billion in new funding at a $12 billion valuation from investors like SoftBank, D1 Capital Partners, and Altimeter Capital.
fundraisingenterprise_aiopenaiventure_capital
News hackernews 6 days ago

Lovable raises $400M Series C

By thoughtpeddler

72 score
AI Analysis

Lovable announces a $400M Series C funding round.

Lovable raises $400M Series C
fundingapplicationsvibe-coding
News The Decoder 6 days ago

Researchers can now reverse-engineer LLM prompts from output text with near-perfect accuracy

By Matthias Bastian

70 score
AI Analysis

Researchers at IIT Bombay and Adobe Research have built an inverse language model called Previous-Token Prediction that reconstructs original LLM prompts from output text with near-perfect accuracy, without access to model weights and across multiple models. The result threatens the confidentiality of proprietary system prompts.

Researchers at IIT Bombay and Adobe Research have built an inverse language model that reconstructs the original prompt from an LLM's output with near-perfect accuracy. Their method, called "Previous-Token Prediction," doesn't need access to model weights and works across different models. For companies relying on proprietary system prompts, this could be a serious security risk. The article Researchers can now reverse-engineer LLM prompts from output text with near-perfect accuracy app
ai_researchai_securityllm_prompt_extraction
News AI News & Artificial Intelligence | TechCrunch 6 days ago

Everything announced at Made by Google ’26: Pixel 11, Pixel Watch 5, Pixel Tag, and tons of Gemini features

By Lauren Forristal

68 score
AI Analysis

At Made by Google 2026, Google unveiled the Pixel 11 lineup, Pixel Watch 5, a new AirTag competitor called Pixel Tag, and a broad set of Gemini-powered features across hardware and software.

From the Pixel 11 series and a brand new competitor to Apple’s AirTag, here are all the announcements from the Made by Google 2026 event.
googlegeminiconsumer_hardwareai_integration
News huggingface.co 5 days ago

Meet North Micro Vision: A 2.4B Native-Resolution Vision-Language Model

68 score
AI Analysis

Cohere Labs released North-Micro-Vision-Instruct, a 2.4B-parameter open-weight vision-language model with native-resolution image support, under Apache 2.0. It combines a 2B language model with a custom-trained 400M native-resolution vision encoder.

Back to Articles Meet North Micro Vision: A 2.4B Native-Resolution Vision-Language Model Community Article Published August 12, 2026 Upvote 26 +20 David Rau d-rau Follow CohereLabs Florian Schneider floschne-cohere Follow CohereLabs All co-authors listed below Today we are releasing North-Micro-Vision-Instruct , a 2.4B-parameter open-weight vision-language model (VLM) with native-resolution image support, under the Apache 2.0 license . North Micro Vision is our smallest VLM to date. It is d
open_source_releasevision_language_modelscoheremultimodal
News Ars Technica - All content 6 days ago

Researchers found a way to hijack devices through Zoom screen sharing

By Lily Hay Newman, wired.com

65 score
AI Analysis

Researchers at A Security disclosed Zoom vulnerabilities exploitable via screen sharing that could silently hijack devices, using publicly available AI models to find and weaponize the bugs in fewer than 20 prompts. Zoom issued a security advisory after disclosure.

As AI models gain advanced capabilities to find vulnerabilities in software, develop ways to exploit them, and even carry out autonomous hacking sprees, researchers offered a sobering new example on Tuesday, disclosing vulnerabilities in the video conferencing platform Zoom that could have been exploited to take over targets’ devices. Anyone on a call that involved screen sharing, whether participants or the host, would have been vulnerable to a silent attack that could be carried out with no in
ai_securityvulnerability_researchai_capabilities
News AI | The Verge 6 days ago

Grok is now an AI ‘teammate’ you can assign work

By Jess Weatherbed

65 score
AI Analysis

SpaceXAI has launched Grok Bot, an always-on AI agent that operates as an autonomous teammate in a shared cloud environment, signing into apps and tools to complete multi-step workplace tasks. The launch pits xAI against OpenAI ChatGPT Work, Anthropic Claude Cowork, and Microsoft Copilot in the enterprise agent race.

You’ll have to be fine with letting Grok sign into your online accounts, however. | Image: SpaceXAI SpaceXAI has introduced Grok Bot, an always-on AI agent service designed to behave like independent "AI teammates" that can do your work for you. The bots share their own cloud-based computer environment, and can sign into apps, tools, and websites you already use to complete multi-step workplace tasks, only coming back when their assigned work is completed or if something requires approva
ai_agentsenterprise_aixai
News The Decoder 6 days ago

Google's Gemini is losing market share to ChatGPT and Claude according to new market data

By Matthias Bastian

65 score
AI Analysis

Three independent data sources (Pangram, Similarweb, OpenRouter) show Google Gemini losing AI market share, dropping from about 12 percent to 1.9 percent on Pangram while OpenAI holds over 50 percent and Anthropic grew from 4.3 to 14.9 percent.

Three data sources tell the same story: Google's Gemini is losing AI market share. Pangram reports a drop from 12 to 1.9 percent, while OpenAI holds over 50 percent, and Anthropic grew from 4.3 to 14.9 percent. Similarweb and OpenRouter confirm the trend. The article Google's Gemini is losing market share to ChatGPT and Claude according to new market data appeared first on The Decoder.
competitive_dynamicsmarket_sharegoogleanthropicopenai