AI Signal Feed
What actually happened in AI today. Ranked by evidence, impact, and actionability. Not engagement.
Cloudflare mandates AI companies to separate web crawlers for search and training
Cloudflare has implemented a policy requiring AI companies to separate web crawlers for search from those used for AI training by September 15, 2026.
Reflection AI signs $150 million monthly deal with SpaceX for Nvidia AI chips
Reflection AI will begin paying $150 million per month for access to Nvidia's GB300 AI chips and hardware at SpaceX's data center.
$1M grant round launched for AI x-risk projects
A $1M grant round has been launched to fund projects aimed at reducing existential risks from AI, with grants ranging from $5k to $50k.
Oracle lays off 21,000 workers due to AI adoption
Oracle reduced its workforce by 21,000 employees, a 12.9% decrease.
AI systems outperform expert humans in persuasive communication
AI systems demonstrated superior persuasive capabilities compared to expert humans in multiple studies, influencing real-world behavior such as charitable donations.
Advocacy organizations can influence LLM values through Wikipedia edits
Advocacy organizations can strategically edit Wikipedia to influence how LLMs discuss specific topics, such as animal welfare.
Meta allows Instagram users to opt out of AI use of their photos
Instagram users with public accounts are now automatically included in AI generations of their content unless they opt out.
T-Mobile files lawsuit against Broadcom over VMware support obligations
T-Mobile has initiated a lawsuit to clarify Broadcom's obligations regarding VMware support.
Amazon Bedrock introduces MiniMax models for AI workloads
Three MiniMax models (M2, M2.1, M2.5) are now available on Amazon Bedrock for various production AI workloads.
Japan's top court rules AI cannot be listed as inventor on patent applications
The ruling establishes that AI cannot be recognized as an inventor in patent applications in Japan.
Google offers Gemini's personalized AI image generation for free to US users
Gemini's personalized AI image generation is now available for free to eligible users in the U.S.
Study reveals flaws in LLM-as-judge safety evaluations due to temperature settings
Identification of structural gaps in evaluation harnesses for LLM-as-judge components affecting reproducibility of safety evaluations.
Deployment of SeedVR2 for video upscaling on Amazon SageMaker AI
SeedVR2 model is now implemented on Amazon SageMaker AI for video upscaling, enhancing video quality from lower resolutions.
Ampersend develops a pay-per-intelligence routing layer using Amazon Bedrock AgentCore Payments
Ampersend launched a new routing layer that allows AI agents to autonomously pay for intelligence services through a single integration with Amazon Bedrock AgentCore Payments.
Release of MonitoringBench, a benchmark for evaluating coding-agent monitors
Introduction of MonitoringBench, a benchmark consisting of 2,644 attack trajectories for evaluating coding-agent monitors.
TIDAL implements policy to stop monetization of AI-generated music
TIDAL's new policy prevents AI-generated music from earning revenue on its platform.
Introduction of ELADO benchmark suite for assessing neural operator architectures in elliptic PDEs
A new benchmark suite called ELADO has been introduced to assess neural operator architectures in the context of elliptic PDEs.
PP-OCRv6 model released on Hugging Face with expanded language support and increased parameters
The PP-OCRv6 model has been launched with support for 50 languages and an increase in parameters from 1.5M to 34.5M.
Samsung Electronics deploys ChatGPT Enterprise and Codex to employees worldwide
ChatGPT Enterprise and Codex have been deployed to Samsung Electronics employees globally.
Introduction of VERITAS, a general-purpose replication tool for scientific research
VERITAS, a new domain-agnostic replication framework, has been developed to automate the verification of scientific research claims.