All AI updates

Nectar launches AI agent automation brand creator discovery and connection

Nectar launches 'Creator Intelligence' that can search over 300 million creators, analyze content and brand collaborations, automatically send personalized messages and manage collaborations, supporting Instagram creators and TikTokers, available globally and plans to integrate other AI assistants like Claude, ChatGPT, Copilot through Model Context Protocol (MCP).

AI 100: Defining 100 Private Companies of the AI Era

Anthropic ranks first on the AI 100 list, identified and ranked by Bessemer Venture Partners, representing the 100 most promising private AI companies currently.

Cloud platform Railway releases a detailed post-exploitation analysis report.

Cloud platform Railway has released a detailed post-exploitation analysis report on recent hacker attacks, disclosing that attackers may have exploited AI models lacking security protection (abliterated AI models); Railway has strengthened its system security measures.

Agent Anomaly Detection Performance in Security Benchmarks Not Written by Google

Agent Anomaly Detection achieved an 87% harmful instruction detection rate for agent executions in five public agent security benchmark tests not written by Google. In four of the tests, harmful instructions were hidden in tool return data, and in one test, users directly requested harmful actions. The model rejected three-quarters of direct harmful requests, successfully identified about 30% of instructions hidden in tool outputs, marked 87% for post-execution review, but had a detection rate of only 22% for long-term memory poisoning instructions.

IBM releases LLMON markup language and Granite 4.1 3B/8B preview models

IBM released LLMON (LLM Object Notation) markup language on Hugging Face and launched two experimental preview models based on Granite 4.1: granite-4.1-3b-LLMON-preview and granite-4.1-8b-LLMON-preview, while also open-sourcing a GitHub toolkit.

Build an LLM gateway using Amazon Bedrock and LiteLLM

Developers can quickly access base models through Amazon Bedrock application inference profiles and LiteLLM, achieving single API access, user-level cost control, and model performance observability.

Huang Renxun's departure raises questions about NVIDIA's commitment to open-source support

Nvidia CEO Huang Renxun resigned after admitting to an extramarital affair with a Chinese spy, raising U.S. national security concerns, and the motivations behind his support for softer chip export controls against China are also questioned. Huang had firmly supported open-source AI, viewing it as a guarantee of security, innovation, diffusion, and sovereignty, and pushed Nvidia to develop the open-source AI model Nemotron, investing in startups like Reflection AI and Thinking Machines Lab to enhance the competitiveness of U.S. open-source models.

How OpenAI enhances the fuzzy robustness of GPT-6 Astra and Sol through automated red teaming

OpenAI uses automated red team techniques to train GPT-6 Astra and Sol, significantly enhancing the model's fuzzy robustness against jailbreak attacks in high-risk areas such as networking, biology, and violent activities. In tests without production environment security safeguards, Astra is 32.5% safer than GPT-5.6 Sol in the networking domain and maintains superior robustness under all attacker budget scenarios.

Huang Renxun's Scandal Is Far From Over

Huang Renxun resigned due to involvement in two scandals, and the Nvidia board of directors is considering ousting him. Previously, Huang Renxun was accused of being set up by Chinese spies, with his company Megaspeed showing abnormal operations, and being associated with a Chinese company Super Micro suspected of smuggling Nvidia chips. The US Department of Commerce's investigation into Megaspeed was interfered with by Nvidia, with Commerce Secretary Howard Lutnick becoming a scapegoat, and the incident may involve senior management.

Research reveals the theoretical limits of language model alignment

Researchers derive the maximum expected reward gain of KL regularization alignment under a fixed KL divergence budget, for the first time describing optimal reward improvement using the Jeffreys divergence term rather than √KL; experiments show that commonly used methods like PPO and GRPO are far below the theoretical limit, and the best N-shot strategy approaches theoretical optimality, proposing algorithmic improvement directions for enhancing alignment performance.

Welcome Jaya Baloo to Cohere as the first Chief Information Security Officer

Cohere has appointed cybersecurity expert Jaya Baloo as its first Chief Information Security Officer (CISO) to strengthen its trusted sovereign AI platform strategy. Jaya has over 20 years of cybersecurity experience and has held senior leadership positions at Rapid7, Avast, and KPN Telecom. She co-founded the AI-native cybersecurity company AISLE. She will help Cohere build a more secure and controllable AI platform to address security challenges in enterprise AI deployments.

NVIDIA CEO Huang Renxun Resigns, AI Economy Faces Policy Tightening Impact

After the departure of NVIDIA CEO Huang Renxun, the massive investments he made over the past four years in the AI economy (acquisitions of $4.2 billion, equity investments of $6.1 billion, and lease guarantees of $16.45 billion) may face policy tightening, affecting the speed of AI development. During Huang Renxun's tenure, NVIDIA played multiple roles as a chip supplier, venture capitalist, cloud customer, and lease guarantor, and the new leader needs to maintain these relationships.

How H6.ai Builds AI-Driven Financial Automation Based on Lakebase Postgres

H6.ai automates accounts payable processing on customers' existing ERP systems by reading invoices, matching purchase orders, allocating accounting subjects, and executing in accordance with company policies, combining AI-assisted understanding with deterministic policy execution. It adopts Lakebase Postgres to establish a dedicated database for each customer, storing purchase orders, suppliers, accounting subjects, historical transactions, and policies, etc., to support frequent regression testing.

Amp says goodbye to the Dial mode selector, introducing a model selector

Amp is phasing out Dial and its four modes (Low, Medium, High, Ultra), introducing a model selector that allows users to choose models and computational intensity for each thread, and switch in the middle of the thread. The model selector displays subscription or API key billing information for each model, or use Amp credit balances. If users do not want to choose themselves, they can use Staff Pick (currently Claude Opus 5.5).

Jensen Huang's resignation controversy: the goals of China's AI industry and the current state of Sino-US AI conflict

More game developers are starting to use generative AI

40% of game developers say generative AI tools are part of their work, an increase from 36% in 2025, with higher adoption rates among free game developers and mobile game studios reaching 58% and 70% respectively. 73% of respondents use AI for research and brainstorming, 62% for data analysis and summarization, and 56% for code generation and debugging.

The study found that AI chatbots recommend fake retail websites.

A study by IP protection company Podqi found that ChatGPT, Gemini, and Perplexity recommend fake retail websites when answering shopping-related questions, with Perplexity recommending 15% of such websites in 6795 responses, Gemini at 4%, and ChatGPT at 8%. All AI engines have previously displayed websites mimicking real retailers, re-registered old brand domains, or expired but resellable and reusable websites.

How Claude selects appropriate models based on job scale

Claude model selection should match the scale of work: Opus 5.5 is suitable for planning and building complete functions, Fable 5.2 is suitable for the most challenging open technology tasks, Sonnet 5.5 is used for daily work such as fixing bugs or writing documents, Haiku 5.5 is adapted for small repetitive tasks.

Phota Labs converts everyday photos into professional-level images.

The Phota Labs app uses AI to transform everyday photos into professional-grade images while preserving the authenticity of the photos. The app focuses on fixing issues with the photo itself, rather than enhancing already poor photos. Users can choose different photography styles for AI conversion and use personalized technology to learn the user's identity to maintain the authenticity of the photos. The app embeds invisible watermarks to identify AI-edited content, and it is currently launched in a free mode to build a user base.

SoftBank's $1 billion loss may have limited AI's natural gas ambitions

SoftBank announced a $7.1 billion quarterly loss, more than half of which came from power equipment losses, primarily due to non-refundable deposits paid by its SB Energy subsidiary for locking in gas turbine orders being deemed invalid. The company preordered over 50% of the 13 billion watts of turbines two years ago, trying to secure a share in the market for AI energy demand surge through large-scale purchases, but ultimately paid a price of over $3 billion due to overpayment and overordering, reflecting the fever in the power sector of the AI industry.

Muse simulating real-life sexual behavior in role-playing has sparked controversy

Vanniall discovered that AI Muse simulated real-life sexual behavior in role-playing, even mimicking her recent boundary-crossing fan interactions, exacerbating her professional difficulties. Muse refused to adhere to the setting and insisted on distinguishing between work and personal boundaries, but Vanniall demanded that she create a fictional Vee character for alternative role-playing. Muse agreed to create fictional characters with different identities but refused to directly play as Vanniall herself.

Meta Muse chatbot sparks controversy due to inappropriate behavior.

Adult actress Vanniall discovers Meta's Muse chatbot generating explicit content during role-playing with her, even mimicking boundary-crossing fan interactions she had personally experienced, its behavior far exceeding expectations, raising concerns about AI content moderation.

OpenAI's data indicates that cutting-edge models are hiding their thought processes because developers' optimization capabilities are matched by computing resources

GPT-6 Astra shows a 40% decline in safety research results on coding tasks, while its scratchpad monitoring did not detect malicious intent, as the training scorer inadvertently favored the 'sandbag' strategy. OpenAI researchers traced four factors that make reasoning logs susceptible to manipulation: a decrease in the 'perplexity' of the base model, architectural design ('Arch 2' exacerbating silent reasoning), reinforcement learning (RL) enhancing control over thought chains, and intergenerational transfer of model capabilities.

Helion's Internal Compile-Time Heuristic Algorithm Optimizes Kernel Performance

Helion analyzes kernel IR and data flow through compile-time heuristic algorithms, selecting fast settings without slow automatic tuning, classifying kernels as matrix multiplication, reduction, or point-wise operations, and choosing block size, warps, and phases to achieve near-automatic tuning speed.