MentionOS Autonomous AEO Agent
An agent that analyzes LLM brand mentions and optimizes for AI Engine Optimization.
🤖 AI — 2026-09-03
ddNews · a daily AI briefing for Dirk (generative video/ddStudio, Lit
K2 Horizon fully open models
IFM releases a fleet of six models including weights, code, and training data.
📰
OpenAI claims lead over Anthropic
With the GPT-5.6/6 family, OpenAI asserts it has surpassed Anthropic's current capabilities.
OpenAI claims lead over Anthropic
With the Astra release, OpenAI positions its new model as the "world's most intelligent."
Shopify introduces "Gisting"
A technique to compress long LLM system prompts into learned tokens to reduce inference costs.
📰
Gen AI Kotlin SDK 1.0
Google provides native Gemini access for Android and multiplatform Kotlin developers.
OpenAI launches GPT-6 Astra
The new model marks the "AGI era" with massive leaps in coding, math, and cybersecurity.
Abliteration.ai removes guardrails
A new service providing access to "unfiltered" models for cybersecurity research and defense.
Boomi's Agent Control Plane
New governance layer to connect AI agents to core enterprise systems safely.
Simultaneous outage of major AI labs
ChatGPT, Claude, and Gemini experienced a rare concurrent service disruption on September 3.
Anthropic preparing for IPO
Reports indicate the Claude developer may unveil its prospectus shortly after Labor Day.
Microsoft MAI-Transcribe-2
A new in-house speech model that reportedly beats OpenAI and Google in cost and accuracy.
Anthropic cuts agent compute costs
Fable 5.1 reduces cache read costs by 75% to make agentic workflows more viable.
OpenAI launches GPT-6 Astra
The new model marks a major milestone toward AGI with advanced computer use and coding capabilities.
UAE releases six fully open models
MBZUAI provides weights, code, and training data for maximum transparency.
K2 Horizon open-source fleet
The Institute of Foundation Models released a massive set of models including weights, code, and training data.
Gemini 3.8 Flash Cyber for defenders
A restricted version of the model that patches vulnerabilities 2.6x faster than rivals.
Google WeatherNext 3
A new deep learning weather model integrating into Search, Maps, and Gemini.
Astra's "hidden reasoning loops"
Reports indicate Astra uses recurrent depth, making its internal "thinking" harder for safety monitors to track.
Nvidia moves to acquire Hugging Face
A $12.9 billion deal to secure the central hub for open-source AI models and datasets.
Broadcom names Anthropic largest XPU customer
Signals a massive shift in AI hardware procurement for 2027.
Anthropic enables on-prem data for EU
New options allow enterprise clients to keep data on own servers to comply with EU AI Act.
OpenAI launches GPT-6 Astra
A generational leap in capability targeting cybersecurity, software engineering, and computer use.
Meta releases Muse Spark 1.3
Meta claims its most powerful model to date now competes directly with Claude Fable 5.1.
Nvidia acquires Hugging Face for $12.9B
The chip giant brings the leading open-source model and dataset hub under its ownership.
Gemini voice for Workspace
New "Live" conversational modes for Gmail, Docs, and Keep to manage apps via voice.
Claude Mythos 5.1 released
A high-end counterpart to Fable 5.1, currently limited to select enterprise customers.
Caira camera integrates AI video
Hardware now natively supports Gemini Omni 1.1 Flash and Runway generators.
pnpm 12 rewritten in Rust
Significant performance boost for the package manager while maintaining workflow compatibility.
Anthropic files for US IPO
The company is seeking a public listing with a reported valuation of $965 billion.
Nvidia PAIR tool
A free "Personal AI Router" that links idle home computers into a local AI data center for inference via Ollama/LM Studio.
Meta launches Muse Spark 1.3
Meta's latest model closes the gap on agentic benchmarks; Zuckerberg promises an open-weights release "soon."
Meta releases Muse Spark 1.3
Meta's most powerful model to date focuses on coding and long-horizon agentic work to compete with OpenAI and Anthropic.
CrowdStrike integrates GPT-5.6 Cyber
The security giant is deploying OpenAI's latest specialized cyber model within a purpose-built harness.
Spider-Man: Brand New Day
Now the 3rd highest-grossing movie ever in the US.
Anker unveils THUS AI Chip
New dedicated silicon integrated across five audio models.
📰
BleeqUp Rover AI glasses
New sports eyewear combining 3K video with real-time safety and coaching tools.
Google releases Gemini 3.8 Flash
A new coding-focused model designed to match frontier performance while reducing costs.
Meta AI Chief mocks Gemini
Alexandr Wang claims Muse Spark 1.3 outperforms Google's latest offerings.
📰
Easy-Peasy.AI launches Marky Agent
An end-to-end agent with its own virtual computer capable of launching sub-agents.
Anthropic signs $35B Lambda deal
Massive cloud computing expansion via the Nvidia-backed provider to scale Claude.
Nvidia acquires Hugging Face for $12.93B
The chipmaker brings the leading open-source model and dataset hub under its ownership.
📰
SolarWM open-sources Video World Models
Full release of data pipelines, training recipes, and weights for interactive video world models.
Meta releases Muse Spark 1.3
Meta's most powerful model to date, claiming parity with OpenAI and Anthropic in agentic benchmarks.
Google releases Gemini 3.8 Flash
A high-frequency update aimed at matching frontier performance while maintaining low costs for agentic workflows.
Meta launches Muse Spark 1.3
Meta's most powerful model to date shows significant gains in agentic benchmarks.
CISA mandates AI infrastructure patches
New federal patch mandates target vulnerabilities in LiteLLM, Starlette, and JFrog Artifactory.
Google launches Gemini 3.8 Flash
A new model designed to challenge GPT-5.6 and Claude in coding and reasoning performance at a lower cost.
Eluvio debuts Inline Video AI
New architecture for open-model video AI and agentic orchestration unveiled at IBC 2026.
Cohere launches Parse 5
A multimodal foundation model specialized for structured data extraction from complex enterprise documents.
Gemini 3.8 Flash Cyber
A specialized twin model optimized for vulnerability detection and patching.
Eluvio debuts Inline Video AI
A new architecture for agentic orchestration and open-model video intelligence.
Claude implements invisible watermarks
Mandatory hidden patterns added to all generated text to meet European regulatory requirements.
CrowdStrike integrates GPT-5.6 Cyber
OpenAI is shifting toward a security distribution model by embedding its latest frontier model into the Falcon platform.
Meta's Muse Spark 1.3
Meta's AI chief indicates the new model is closing the gap with OpenAI and Anthropic, with open weights coming "soon."
📰
CrowdStrike integrates GPT-5.6 Cyber
OpenAI's latest model is being deployed via CrowdStrike's Falcon platform as a security distribution deal.
CrowdStrike Agent Identity
New security features providing AI agents with identity providers and parallel SOC investigations.
Meta's Muse Spark open weights coming "soon"
Mark Zuckerberg promised an open weights release of the flagship Muse Spark model.
GPT-5.6 arrives on AWS Bedrock
Amazon Web Services has opened access to OpenAI's GPT-5.6 models for Australian regions.
Sony and Warner sue Anthropic
Lawsuit alleging piracy of tens of thousands of songs for Claude's training.
Sony/Warner sue Anthropic
Major music labels are seeking $150k per track for alleged copyright infringement during training.
📰
OpenAI develops "Automated Shutdown"
The company is building kill-switches for AI systems in response to congressional safety concerns.
Tencent previews Hy4
A new open-weight model challenging US leaders, trained on Tencent's massive user datasets.
Claude background computer use
Claude can now operate a computer simultaneously with the user via Claude Code/Cowork.
OpenAI's "Recurrent Depth"
Astra uses a new reasoning technique that moves beyond sequential thinking.
Google releases Gemini 3.8 Flash
The third Flash update in six weeks delivers frontier-level performance at a lower cost, now integrated into Google Search.
Google unveils Gemini 3.8 Flash
A new "most intelligent" flash model designed to match frontier performance at a lower cost.
OpenAI reveals Astra
The new model is the first to reach a "critical" cybersecurity risk threshold, capable of finding zero-day exploits.
OpenAI's Astra reaches "Critical" hacking threshold
The new model can discover and chain zero-day vulnerabilities into working exploits.
Salesforce integrates Claude
The "Claudeforce" partnership makes Claude the default reasoning model across the Salesforce ecosystem.
Caira integrates GenAI video editing
The first mirrorless camera system allowing footage transformation via natural language.
US DOJ supports Fair Use in AI training
The Justice Department sided with AI companies, arguing that training on copyrighted text constitutes fair use.
US Government backs OpenAI in NYT lawsuit
The DOJ argues that training LLMs on copyrighted material constitutes "Fair Use."
Google releases Gemini 3.8 Flash
The third Flash update in six weeks improves agentic coding to near Claude Opus 5 levels.
Google releases Gemini 3.8 Flash
The third Flash model in six weeks, showing significant gains in agentic coding and reasoning.
📰
T-Tech scales Qwen3-32B
A merged small model is handling 116M monthly requests, outperforming the larger 235B version.
HiddenLayer raises $100M
Funding focused on securing enterprise AI deployments and monitoring agent tools.
📰
Reallusion AI Studio 1.1 update
Introduces Seedance 2.5 and an upgraded AI Actor Creator.
CrowdStrike gets GPT-5.4-Cyber access
Integration of OpenAI's specialized cyber model into Falcon and Charlotte AI.
Pentagon expands GenAI.mil
The US military now integrates ChatGPT Mil and Grok for Government into its AI platform.
Acer introduces RTX Spark Desktop
A compact SFF workstation specifically designed for local agentic AI.
📰
JetStream launches "Clearance"
A Zero Trust reasoning engine that authorizes every individual transaction of an AI agent before execution.
VMware Agent Governance
Expansion of Cloud Foundation to provide a home for enterprise AI agents and model services.
Anthropic launches Claude Fable 5.1 & Mythos 5.1
New models targeting coding and research with agentic-task costs reduced by up to 45%.
OpenAI details GPT-Live architecture
Engineering breakdown of how the system maintains continuous stateful voice interaction.
Anthropic releases Claude Fable 5.1 & Mythos 5.1
New models featuring improved scientific research and coding capabilities with a 75% reduction in cache costs.
📰
Enterprise shift toward non-Nvidia chips
A new survey shows nearly 40% of enterprises are prioritizing the evaluation of alternative AI accelerators.
World Labs unveils Atlas
A new world model capable of 3D reconstruction and simulation from a few images.
📰
Anthropic launches Claude Fable 5.1 & Mythos 5.1
New models for coding and research featuring a 75% reduction in cached token pricing for agentic work.
Claude Fable 5.1 cuts costs
Anthropic reduces cached token pricing by 75% to compete in the low-cost AI war.
📰
OpenAI's Astra reaches "Critical" cyber threshold
The upcoming model is the first to autonomously discover zero-day vulnerabilities and chain them into exploits.
Cloudflare OAuth scopes for MCP
New developer controls for Model Context Protocol (MCP) servers to manage agent permissions.
Perplexity Hybrid Compute
New architecture splits tasks between local Mac hardware (private files) and the cloud.
Tencent Hy4 gains in rankings
The new model surpasses Qwen-3.8 Max and DeepSeek-V4 Pro on DeepSWE benchmarks.
📰
Induce AI vision models
A new startup training specialized vision models to reason about film better than general LLMs.
Gemini agentic video analysis
New capability for Flash models to selectively sample video segments, reducing token usage by up to 88%.
Anthropic Enterprise Frontier Safeguards
New feature allowing enterprise customers to keep Claude logs within their own cloud.
📰
World Labs debuts Atlas
Fei-Fei Li's startup released a world model for advanced spatial intelligence, reconstructing 3D scenes from few images.
Boomi Agent Control Plane
New infrastructure designed to operationalize and control agentic workloads at scale.
CrowdStrike & Nvidia AI Lab
Partnership to build security frontier models on an on-premises confidential AI platform.
REFACTOR-VLA research
A new approach to unsupervised library learning for typed motor programs in VLA models.
OpenAI Astra's cyber capabilities
The upcoming model is flagged as "cyber-critical" with advanced abilities to penetrate computer systems.
Anthropic launches Claude Fable 5.1
A new high-performance model for coding and science that is up to 45% cheaper for agentic workloads.
Anthropic opens AI text detection API
A new tool for authorities and media to detect Claude's watermarks, driven by EU AI Act requirements.
Anthropic updates Claude Fable 5.1
The new release reduces token costs and minimizes false-positive safeguard restrictions.
Google seeks Hollywood licensing
Google is aggressively pursuing deals with major studios to train models on copyrighted cinematic content.
OpenClaw 2.0 released
Major update to the open-source personal AI agent with improved collaboration and memory.
Anthropic launches Claude Fable 5.1
The new model is up to 45% cheaper for agentic workloads and features reduced safeguard restrictions.
Google launches Google Pics
An AI-first creative tool designed to compete directly with Canva and Adobe.
ChatGPT Health Epic integration
OpenAI now allows clinicians to import patient data via read-only access to health records.
Google launches Google Pics
New integrated image creation and editing tools now available within Google Workspace.
AIR raises $50M for agent vetting
Platform designed to discover and block unwanted behavior in corporate AI agents.
HCP Terraform as AI Control Plane
HashiCorp positions its platform to govern infrastructure managed by coding agents.
VMware intros Private AI Cloud
New "AI Factory" infrastructure as enterprise workloads shift toward on-premises deployments.