Google, Microsoft, OpenAI, and Anthropic are backing a new standard that lets website owners opt out of having their content used for AI model training. The companies plan to implement a common signal at the Internet Engineering Task Force that applies across multiple AI crawlers. Anthropic has already adopted the signal, while Google and Microsoft plan to support it next year and OpenAI says it will implement it in the coming months. Microsoft emphasized that publishers should not have to choose between visibility on search engines and controlling how their work is used for AI training.
Etsy updated its Terms of Use and Privacy Policy to explicitly allow the use of seller content for training AI models. The company already had a license to use user content but has now added AI training permission. Etsy also tightened rules around third-party use of marketplace data and AI shopping agents, while offering opt-out rights described in its Privacy Policy.
Amazon and Intel shook up tech markets on Friday, deepening investor concerns about the cost and return timeline of AI spending. Nasdaq futures fell more than 2% after Amazon signaled continued heavy investment and Intel issued a warning. European shares also slipped, with the STOXX 600 index down 0.29% to 647.23 points. Technology stocks led the declines, including ASML down 5% and other chip makers falling between 3.3% and 5.7%.
NVIDIA launched NemoClaw, a blueprint system for building specialized autonomous AI agents. It works with three agent harnesses: OpenClaw, Hermes Agent, and LangChain Deep Agents. The system includes OpenShell for security and policy enforcement, and NeMo Relay for session tracing and audit trails. NemoClaw agents can run in the cloud, data centers, on NVIDIA DGX Spark, and on NVIDIA RTX PCs and workstations.
Researchers introduced SkillCT, a benchmark for evaluating AI agent skills across nine dimensions. Unlike existing benchmarks that focus mainly on task execution, SkillCT assesses specification quality, behavioral robustness, semantic alignment, and responsibility boundaries. It includes an execution-free baseline that predicts quality profiles without running the skills, reducing evaluation time and token use. SkillCT-guided optimization was shown to improve task success and efficiency.
Coveo acquired Anglera, a San Francisco-based AI company backed by Y Combinator. Anglera uses agentic AI to extract and standardize product attributes from sources like PDFs and images. A 30-day test showed enriched product pages generated a 120% increase in organic traffic and a 270% increase in AI-referred visits. Coveo plans to integrate Anglera's capabilities into its Commerce platform.
Apple acquired AI firm Ascend.io in a deal that gives Apple the right to hire certain employees and license the firm's intellectual property. Ascend provides software used to automate and manage data pipelines and pioneered Agentic Data Engineering over the past 10 years. The acquisition gives Apple access to Ascend's patents and technology talent.
Key Takeaways
- Google, Microsoft, OpenAI, and Anthropic support a new standard letting publishers opt out of AI model training
- Anthropic has already adopted the opt-out signal; Google and Microsoft plan support next year; OpenAI will implement it in coming months
- Etsy updated its policies to allow AI training on seller content while offering opt-out rights
- Nasdaq futures fell over 2% as Amazon and Intel deepened AI spending concerns
- European STOXX 600 index dropped 0.29% to 647.23 points amid tech stock declines
- NVIDIA launched NemoClaw, a blueprint system for building specialized autonomous AI agents
- SkillCT benchmark evaluates AI agent skills across nine dimensions including specification quality and behavioral robustness
- Coveo acquired Anglera, which uses agentic AI to extract product attributes from PDFs and images
- A 30-day test of Anglera's enriched product pages showed a 120% increase in organic traffic and 270% increase in AI-referred visits
- Apple acquired Ascend.io for talent and patents in Agentic Data Engineering
Major AI companies to offer simpler training control tool for publishers
Google, Microsoft, OpenAI, and Anthropic plan to adopt a new proposal at the Internet Engineering Task Force. The tool will let web publishers choose how their content is used for AI training. It aims to give creators clearer options while keeping their work searchable on Bing. Microsoft stated that publishers should not have to pick between visibility and controlling AI use.
AI giants agree to let websites opt out of model training
Anthropic, Google, Microsoft, and OpenAI agreed to support a system that lets website owners opt out of AI model training. The companies will implement a common signal for expressing content use preferences. The change comes after these firms trained models on massive amounts of online content. OpenAI said it will honor the signal for foundational generative AI training.
Top AI firms back new standard to opt out of scraping
OpenAI, Google, Microsoft, and Anthropic are supporting a new standard that lets publishers opt out of AI scraping. The control will apply across multiple AI crawlers. Anthropic has already adopted the signal. Google and Microsoft plan to support it next year, and OpenAI says it will implement it in the coming months. Publishers have struggled to stop unauthorized scraping of their sites.
LinguaArray uses LLMs to guide formal SMT array solving
Researchers created LinguaArray, a framework where large language models suggest key terms for solving quantified array problems. A traditional SMT solver remains the only proof engine and verifies all results. The LLM proposes a small Semantic Focus Set of ground terms and constraints. Testing on 1015 SMT-LIB instances showed the system solved up to 69.1% of cases and improved solved-instance counts by 105% to 397% over backend solvers alone.
Study questions LLMs causal reasoning on benchmark tests
A new study examines whether large language models truly perform causal reasoning on the CLadder benchmark. Researchers found that the benchmark contains superficial patterns that models can exploit. Even simple models like random forests perform well on CLadder data after training on generated samples. Removing key text sections had almost no effect on LLM performance. The study suggests strong results come from pattern matching rather than deep causal understanding.
Etsy Updates Rules to Allow AI Training on Seller Content
Etsy changed its Terms of Use and Privacy Policy to allow itself and approved partners to use seller content for training AI models. The company already had a license to use user content but now adds explicit AI training permission. Etsy also tightened rules for third-party use of marketplace data and AI shopping agents. The new policy includes opt-out rights described in its Privacy Policy.
European Shares Slip as Tech Stocks Fall on AI Spending Worries
European shares edged lower on Friday as technology stocks dropped amid concerns over rising AI investment costs. The pan-European STOXX 600 index fell 0.29% to 647.23 points but was still set for weekly and monthly gains. Amazon and Intel earnings reignited worries about heavy AI spending and returns. Technology stocks led declines with ASML down 5% and other chip makers falling between 3.3% and 5.7%. French inflation and political concerns also weighed on sentiment, but strong earnings from ASML, Shell, and Airbus supported confidence.
Nasdaq Futures Fall as Amazon and Intel Deepen AI Spending Fears
Nasdaq futures slid more than 2% on Friday after Amazon announced plans for continued heavy AI spending and Intel issued a warning. The move reflected growing investor concern about the cost of AI investment and the timeline for returns. The drop came as technology stocks faced pressure across European and US markets. Concerns about AI spending intensity and profitability rattled investors.
NVIDIA NemoClaw Helps Developers Build Specialized AI Agents
NVIDIA launched NemoClaw, a blueprint system for building specialized autonomous AI agents on an open stack developers control. NemoClaw works with three agent harnesses: OpenClaw, Hermes Agent, and LangChain Deep Agents. It combines model and harness optimization with runtime governance and observability tools. The system includes OpenShell for security and policy enforcement, and NeMo Relay for session tracing and audit trails. NemoClaw agents can run in the cloud, data centers, on NVIDIA DGX Spark, and on NVIDIA RTX PCs and workstations.
SkillCT Introduces New Benchmark for Assessing AI Agent Skills
Researchers introduced SkillCT, a benchmark for evaluating AI agent skills across nine dimensions. Current benchmarks focus mainly on task execution, but SkillCT assesses specification quality, behavioral robustness, semantic alignment, and responsibility boundaries. It evaluates skills through static, dynamic, and human-assisted assessment. The framework also includes an execution-free baseline that predicts quality profiles without running the skills, reducing evaluation time and token use. SkillCT-guided optimization was shown to improve task success and efficiency.
Coveo buys Anglera to boost AI product data for e-commerce
Coveo acquired Anglera, a San Francisco-based AI company backed by Y Combinator. Anglera uses agentic AI to extract and standardize product attributes from sources like PDFs and images. The deal brings Anglera's enterprise customers, including multiple Fortune 500 companies, into Coveo. A 30-day test showed enriched product pages generated a 120% increase in organic traffic and a 270% increase in AI-referred visits. Coveo plans to integrate Anglera's capabilities into its Commerce platform.
Apple acquires AI firm Ascend.io for talent and patents
Apple acquired AI firm Ascend.io in a deal that gives Apple the right to hire certain employees and license the firm's intellectual property. Ascend provides software used to automate and manage data pipelines. Over the past 10 years, Ascend developed declarative data pipeline technology and pioneered Agentic Data Engineering. The acquisition gives Apple access to Ascend's patents and technology talent.
Sources
- Google, Microsoft, OpenAI, Anthropic to offer streamlined AI training tool | MLex | Specialist news and analysis on legal risk and regulation
- Anthropic, Google, Microsoft, and OpenAI Agree to Let Sites Opt Out of AI Training
- The biggest names in AI are backing a new way to opt out of scraping
- Verifiable LLM-Guided Focal SMT Solving for Quantified Arrays
- Evaluating Causal Reasoning in LLMs: A Case Study in the Challenges of Procedural Generation
- Etsy Expands AI Training Rights Over Seller Content, Tightens Rules For Third-Party AI Use
- European shares ease as AI-driven tech selloff deepens
- Nasdaq futures sink as Amazon, Intel intensify AI spending fears
- Build Specialized Agents on a Stack You Control with NVIDIA NemoClaw
- SkillCT: A Multidimensional Benchmark and Execution-Free Baseline for Agent Skill Assessment
- Coveo Acquires Anglera To Expand AI-Powered E-Commerce Product Data
- Apple Acquires AI Firm Ascend.io Talent and Patents
Comments
Please log in to post a comment.