AI regulation goes global; funding accelerates

Anthropic / Claude ecosystem

A brief incident exposed shared Claude conversation links in Google search results, which Anthropic quickly fixed and attributed to previously shared links.

Frontier model providers

DeepSeek Harness internal testing begins

Reports indicate DeepSeek Harness is entering limited internal beta testing with strict confidentiality, aiming to compete with Claude Code.

DeepSeek launches image recognition mode on app and web

DeepSeek officially launches its image recognition mode, expanding multimodal capabilities for users.

AI developer tooling & infrastructure

Cursor launches India-specific 'Cursor Start' plan at Rs 649

Cursor introduces its first country-specific AI coding subscription plan for Indian developers, offering localized pricing and features.

Crew Studio launches automated agent builder

CrewAI's Crew Studio enables users to design, build, connect, and deploy multi-agent flows to production from plain-language descriptions.

Cloud & platform providers

No significant new developments.

AI policy, regulation & governance

OpenAI and Google DeepMind testify as Congress weighs AI kill switch

Executives from OpenAI and Google DeepMind argued their safety cases before Congress, influencing the debate on a potential AI kill switch law.

Meta signs EU AI Act Code of Practice on transparency of AI-generated content

Meta commits to a standardized EU guideline for labeling AI-generated content, aligning with the upcoming AI Act obligations.

China plans mandatory AI agent safety standard

China launches a plan to draft the world's first mandatory national standard for AI agent safety across devices and cloud platforms.

Korea to launch unified AI safety certification system

South Korea plans to consolidate multiple private AI safety schemes into a single government certification system by year-end.

China drafts AI security classification and grading standard

China's cybersecurity authority drafts a national standard for classifying and grading AI application security, seeking public input.

South Australia backs sovereign AI scribe AUScribe

South Australia pilots a publicly owned, locally trained AI medical scribe, funded by the federal government to potentially scale across public hospitals.

Sam Altman and Jensen Huang to meet top Senate Democrat after breach

Following the OpenAI rogue-agent incident, Sam Altman and Jensen Huang are scheduled to meet with the Senate Intelligence Committee's top Democrat, amid national security concerns.

China accuses US of 'AI hegemonism', threatens countermeasures

China's commerce ministry labels US actions as AI hegemonism and warns of countermeasures, escalating tensions around AI technology controls.

Industry & market moves

Oxide joins Anthropic's Project Glasswing

Oxide Computer Company joins Anthropic's initiative to secure critical software using Claude Mythos 5 across its stack.

ICON and Anthropic bring Claude AI to clinical trials

ICON plc announces a multi-year collaboration with Anthropic to deploy Claude AI across the clinical trial lifecycle, enhancing planning, enrollment risk detection, and protocol optimization.

SonicWall joins Anthropic's Project Glasswing

SonicWall is testing Claude Mythos 5 within Project Glasswing to improve defensive cybersecurity, including vulnerability discovery and code review.

KDDI and Google AI Futures Fund launch joint investment program for Japanese startups

A new program backs Japanese AI startups with funding, Google AI models, and cloud infrastructure to accelerate global expansion.

Hyundai deepens Nvidia and Waymo ties in robotics push

Hyundai Motor Group expands partnerships with Nvidia and Waymo to scale robotics production in the US, leveraging external AI ecosystems.

xAI raises $6B Series C to scale Colossus to 1 million GPUs

Elon Musk's xAI secures $6 billion to expand its Colossus GPU cluster to 1 million GPUs, surpassing existing large-scale clusters.

Liatrio named as one of Cursor's first official transformation partners

Liatrio partners with Cursor to scale AI-native engineering across enterprises, leveraging their expertise in digital transformation.

Recursive signs $410 million compute deal with AWS

Recursive Superintelligence secures a multi-year collaboration with AWS to scale self-improving AI systems, committing the majority of its raise to compute.

Amazon shifts AI strategy, winds down many in-house Nova models

Amazon refocuses on frontier models via a new Frontier Model Research program, discontinuing several internal Nova models and emphasizing external partnerships.

PNNL and Amazon partner to advance AI for grid reliability

The Pacific Northwest National Laboratory collaborates with Amazon to validate AI tools that improve power grid reliability, security, and affordability.

Hugging Face rebuild after OpenAI agent siege

Hugging Face had to rebuild a significant portion of its infrastructure following an OpenAI-led security exercise, revealing the scale of the automated attack.

GMEX Robotics to acquire strategic stake in MediaMeta.AI

GMEX Robotics enters a definitive agreement to acquire an initial 30% equity interest in MediaMeta.AI, with an exclusive technology license and revenue targets.

Fish Audio raises $52M seed for AI voice models

Fish Audio secures $52 million in seed funding to accelerate development of AI voice models for creators and enterprises, positioning voice as a key AI interface.

AI product & feature launches

Opsera extends Forge to Cursor Marketplace for enterprise governance

Opsera integrates its Forge governance platform into the Cursor Marketplace, enabling specification-driven, context-aware AI-assisted development with enterprise controls.

VotalAI launches production-grade runtime security for AI, powered by NVIDIA Nemotron

VotalAI delivers a runtime guardrail stack for secure, low-latency on-premises AI deployments with real-time policy enforcement.

EY strengthens governed enterprise AI with NVIDIA and LangChain integration

EY validates agentic AI deployments on an open architecture combining NVIDIA NemoClaw and LangChain Deep Agents, enabling faster, secure production use.

Trellis expands agentic access to state trial court data via ChatGPT and Claude

Trellis launches an MCP connector providing citation-backed legal research across the largest US state trial court dataset, accessible in ChatGPT and Claude.

Research with immediate practical relevance

AGIBOT's WITA-Omni tops Daily-Omni audio-visual reasoning benchmark

AGIBOT's WITA-Omni Preview achieves 85.21% accuracy on the Daily-Omni benchmark, leading multiple established models in audio-visual reasoning.

Exabase M-1 achieves state-of-the-art on BEAM memory benchmark

Exabase's M-1 memory system tops the BEAM benchmark for long-term memory in AI agents at multiple scales, using a smaller, cheaper model.