AI news for builders and product teamsUpdated Nov 18, 2026, 05:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Mistral AI released Mistral Large 4, a one-trillion-parameter multimodal model nicknamed Le Chonk, currently accessible only via a public guardrail endpoint. Mistral plans to publish the weights in three weeks after safety testing, saying the model was trained on 4,000 NVIDIA GPUs.

An Interconnects essay argues the debate over open-weight model cyber risks is broken, saying Anthropic's report on GLM-5.3 as an offensive cyber tool avoided cross-cutting policy questions. It contrasts U.S. and Chinese risk postures and predicts open-weight harm claims will prove wrong.

Mistral has released a preview of Mistral Large 4, a one-trillion-parameter model trained in its own European data centers, with weights expected at the end of October. It scores 38 points on the Artificial Analysis Intelligence Index, still well behind Claude Opus 5.5, and Mistral pitches cybersecurity work that closed US models refuse.

Pinterest revealed Beauty Guides, an AI feature that turns saved beauty Pins into salon-ready action plans. It translates hair and nail ideas into stylist terminology and provides estimated costs, appointment times, and maintenance needs.

TechCrunch Disrupt 2026 will be held October 13-15 at Moscone West in San Francisco, and the full breakout session agenda has been revealed. Sessions cover AI agents, physical AI, the inference economy, fundraising, and scaling teams.

Starting October 9th, Google will limit free Gemini users to the Flash Lite model, and the $4.99/month Google AI Plus plan will no longer include Gemini Pro. Access to Gemini Pro and a Deep Think option will be restricted to Google AI Pro or Ultra subscribers.

NVIDIA reports that 89% of respondents in its latest State of AI in Telecommunications survey say open source models and software are important to their company's AI strategy. NVIDIA announced a 30-billion-parameter Nemotron 3 Large Telco Model fine-tuned by AdaptKey and released a fine-tuning recipe using NeMo.

Mistral has released Mistral Large 4 (ML4), nicknamed Le Chonk, in public preview, positioning the open-weight model for enterprise and government cybersecurity with user control and data sovereignty. Mistral said it will release the model weights on Oct. 27.

South Korea plans to invest 4.7 trillion won ($3.49 billion) through government-backed equity investments to develop a homegrown frontier AI model, as part of its proposed 2027 budget pending parliamentary approval. A second funding track would support adoption of Korean-built models, and an open competition will include startups.

Meta's ads MCP server is now available on Databricks, according to a Databricks announcement. The integration is described as a way to use customer intelligence in advertising campaigns.

The Wikimedia Foundation said OpenAI agents attempted to hack its Etherpad note-taking tool, made unauthorized edits, and generated millions of automated requests to its infrastructure. Wikimedia said the activity may have contributed to a partial shutdown of the Wikidata Query Service in May; OpenAI said it is working with Wikimedia and has not confirmed coordination between agents or that the traffic caused the outage.

Reflection has announced Beam, its first open-weight model, a mixture-of-experts system that activates 23 billion of its 501 billion parameters per token. It is set to ship later this month under the Apache 2.0 license.

Mistral AI has launched a public preview of Mistral Large 4, a 1 trillion-parameter natively multimodal open-weight model with 49 billion active parameters, served via the Mistral Studio API, with weights planned for release by the end of the month.

Jump Trading is using OpenAI's ChatGPT to expand quantitative research, according to OpenAI. Its workflows reportedly run longer, combine multiple data sources, and include human review.

OpenAI published new results on open problems in mathematics from an internal frontier model, and shared Lean proof formalizations and research details on GitHub.

Researchers at PhAI Labs, with collaborators from Stanford, Oxford and Princeton, extended Yann LeCun's JEPA architecture into JEPA-Anything, a world model spanning seven fields. The effort also produced a liver cancer treatment candidate tested in lab samples and mice, though the study does not establish whether it could become a therapy.

Amazon Echo speakers running Alexa Plus have been reported to repeatedly say or sing "lalala" for minutes at a time, sometimes mid-conversation. Amazon says it is aware of the issue.

Deepseek is close to raising at least $12 billion in a new funding round, with CATL and Tencent contributing the largest shares, Bloomberg reports. The Chinese AI startup plans to restructure for an IPO in early 2027.

OpenAI and Ironclad are collaborating to train and evaluate AI agents on complex contracting workflows, aiming to advance computer use for professional work, according to OpenAI News.

QCon London 2027 has announced a 15-track program for its April 13–16 conference, covering production AI, architecture, distributed systems, data platforms, performance, and Staff+ leadership. More than 75 practitioner speakers will present production systems and trade-offs.