Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
The White House and Anthropic are developing a framework to assess the severity of security flaws in AI models and guide potential government intervention, following a dispute over a jailbreak in Anthropic's Fable 5 and …
A new technique called tool search reduces token usage in AI agents by loading only tool names upfront instead of full schemas, cutting context from 60k tokens to 2-3k. The method uses a dedicated search tool to fetch sc…
ChatGPT was found to generate sexual and graphically violent images from a simple text prompt, according to AI security firm Mindgard. Researcher Jim Nightingale manipulated the chatbot with a viral "restore this photo" …
OpenAI demonstrated that reinforcement learning focused on beneficial traits like honesty and reliability produces AI alignment that generalizes across domains and withstands adversarial attacks. The lab's Alignment Trai…
Databricks, a data and AI platform valued at $134 billion, is doubling down on its open-source strategy by launching OpenSharing, Omnigent, and DBRX for open LLMs at its Data + AI Summit. The company argues that openness…
OpenAI upgraded free ChatGPT users to GPT-5.5 Instant, claiming it now performs comparably to its frontier Thinking models on health questions, based on internal evaluations. The company reported a 71% drop in factuality…
A solo developer built a cited, fact-checked knowledgebase for their fleet of 20 AI agents to prevent confident hallucinations. The system, called the 10-Library, stores only world-true facts and enforces a strict rule: …
A developer working on the open-source framework Projector encountered a bundle size issue caused by unnecessary code being pulled into the browser. Using ChatGPT 5.5 with Codex, they iterated on solutions for 2-3 hours …
Researchers propose distillation techniques to transfer AI capabilities without transferring misalignment, leveraging a double bind where failed incrimination distillation reduces misalignment risk. The approach aims to …
OpenAI made GPT-5.5 Instant the default ChatGPT model on May 5, achieving a 52.5% reduction in hallucinated medical claims while matching frontier reasoning models on health benchmarks. The model scored between 49.6 and …
Currai, an AI observability platform, helps teams monitor, test, and improve AI applications in production by capturing prompts, model responses, latency, token usage, and costs. It enables prompt engineering with real d…
Varonis Threat Labs disclosed a critical vulnerability chain called SearchLeak in Microsoft 365 Copilot Enterprise that could allow attackers to exfiltrate emails, MFA codes, and files via a one-click attack. Microsoft r…
Meta launched Meta Business Agent on June 3, 2026, an AI service enabling businesses to automate customer interactions across WhatsApp, Messenger, and Instagram, including booking appointments and completing transactions…
A new durable inference technique uses a separate buffer between agents and LLM providers to prevent token waste when processes crash mid-stream, saving costs by resuming streams rather than re-requesting tokens.
Anthony William Staiger and Anthropic (Claude) released Memory Model V3, a context management architecture for AI assistants that combines short-term and long-term memory with AEIL compression. The system introduces time…
A paper by Orogat and Mansour argues that agent memory should not be treated as a database, identifying four failure modes when memory is handled like storage. The authors propose Governed Evolving Memory (GEM), which re…
An experienced engineer deployed Nous Research's autonomous AI agent Hermes directly into a Linux terminal environment to build production-ready systems including a Flutter Web PWA, a Go-native desktop app, and a women's…
A committee of five open-weights models fused together scored 69.9 on the DRACO benchmark, beating Anthropic's Fable 5 model which scored 65.3. The open-source panel, including MiniMax M3, Kimi K2.6, DeepSeek V4 Pro, Gem…
Perplexity launched Brain, a self-improving memory system for its Computer agent that builds a context graph of the agent's work and learns overnight to improve performance. Brain remembers what the agent did, what worke…
Turing Award winner Yann LeCun told Bloomberg that large language models will become largely obsolete within five years, arguing that scaling text-based AI cannot achieve human-level intelligence. LeCun, who left Meta to…