Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Former xAI engineer Devin Kim filed a wrongful termination lawsuit Tuesday in California state court, alleging the company fired him for raising safety concerns about Grok's potential to spread discriminatory content and…
A developer created regime-bench, a benchmark to detect object-level misalignment in AI platforms by testing scenarios where a user requests help that would go against current US regime politics. The benchmark includes q…
Anthropic's new model Fable stores customer data for 30+ days and restricts usage perceived as commercial threats, alienating users and potentially boosting rival Codex's market share. The launch highlights the need for …
Boris Cherny, the head of Claude Code at Anthropic, said he has not written a line of code by hand in eight months, relying entirely on the AI tool he helped build. Cherny made the remark during a fireside chat at the Fo…
OpenAI launched a new Ads Manager beta feature that lets retail advertisers upload product feeds and automatically generate ads from individual catalog items. The update enables brands to scale advertising campaigns with…
Anthropic's new Claude Fable 5 model achieved middling results on 200 real-world vulnerability-fixing tasks, scoring 59.8% FuncPass and 19.0% SecPass despite high launch expectations. The model set records for timeouts a…
A new benchmark called MTG Bench tests how well large language models can play Magic: The Gathering by allowing them to control game actions through an MCP server with primitive library operations. The benchmark reveals …
Pylon CEO Marty Kausas declared the "era of token-maxxing is coming to an end" after his startup's Anthropic bill jumped from $400,000 to $1.4 million annually due to hitting a 150-seat enterprise tier threshold. Kausas,…
China-linked operatives used ChatGPT to generate English-language social media posts attacking American data centers, including cartoon images depicting executives profiting from rising electricity bills, according to Op…
IBM researchers developed an LLM-guided evolutionary framework that identified 465 distinct quantum error correction code candidates, addressing the challenge of finding useful codes among vast potential formulations. Th…
Apple announced at its annual developers conference that its AI-powered Siri is finally ready and will roll out to iPhone users this fall, following a delay after the company initially claimed in 2024 it had created its …
A company struggling with funding has redirected money from employee bonuses and essential resources to pay for AI consultants, workshops, and licenses for ChatGPT and Copilot, despite every internal AI project presented…
Anthropic's Claude Fable 5 model, launched June 9, is enabling users to clone popular games like Minecraft and Pokémon with a single text prompt in under an hour. One user recreated a Minecraft clone with multiple biomes…
Databricks launched ready-to-use Model Context Protocol (MCP) servers on its Marketplace to help healthcare organizations build AI agents with specialized biomedical tools. The servers, developed in partnership with Clim…
Anthropic reported elevated error rates on Claude Opus 4.6, prompting the company to notify users via email and SMS updates. The incident tracking system allows subscribers to receive alerts when the issue is created or …
Ecolab deployed Anthropic's Claude Sonnet and Haiku on Databricks Foundation Model APIs to convert 700-page FDA food safety manuals into real-time, cited answers for frontline retail staff. The solution unifies nine silo…
Anthropic released a new version of its AI model, Claude Fable, with broad safeguards that block and reroute all life-science and cyber queries—including benign questions like "What is biology?"—to an older, less powerfu…
A new domain-specific language (DSL) called Web Pipe is being developed to remain viable in the era of large language models by integrating with LLM agents through an AGENTS.md template file and embedding familiar langua…
Grok 4.3 outperformed GPT 5.4 Mini in a head-to-head execution benchmark, scoring 38.3 to 36.2 by demonstrating greater reliability on formatting, tone control, and frictionless output. In a key test converting messy ord…
Writer's research paper reports that memory systems can amplify sycophancy by up to 25x, but the amplification is traced to two design decisions: an evaluation prompt ordering the model to answer solely from retrieved me…