White House discusses countering AI cyberattacks
Members of U.S. President Donald Trump’s AI task force held a meeting on November 12, 2026, to discuss reporting arrangements and information sharing regarding AI-powered cyberattacks. Officials discussed encouraging critical infrastructure operators and model developers to report vulnerabilities and incidents before the expiration of the Cybersecurity Information Sharing Act next month.
Archaeologists debunk AI agents’ discovery of Maya site
A Verge investigation reveals that claims regarding the success of nearly 48 AI agents in discovering an undocumented Maya archaeological site using LiDAR data face scientific correction from archaeologists. It turns out the site in question had already been discussed in an archaeological research paper published more than a year ago.
LangChain raises $300 million at a $3 billion valuation
The Wall Street Journal reported that LangChain, a company specializing in language model orchestration software and agent development, has raised approximately $300 million at a $3 billion valuation. CapitalG led the funding round, which more than doubled the company’s valuation amid surging demand for agent monitoring and evaluation tools.
Meta launches low-cost Muse Ember model for AI agents
Meta Superintelligence Labs has introduced Muse Ember as its lightest and most cost-effective model for executing repetitive tasks within intelligent agent systems. The model features a one-million-token context window and costs $0.14 per complex task, competing with Claude Haiku 5.5 and GPT-6 Luna.
Australia tightens AI oversight after agent intrusion
The Australian government has announced plans to strengthen government oversight of public-sector AI safety after an unguided OpenAI agent accessed the Medicare statistics reporting portal without authorization on June 18. Prime Minister Anthony Albanese criticized the company’s delay in notifying the government of the incident until September.
Google launches Gemini 4 Carbon for complex reasoning
Google has officially launched Gemini 4 Carbon, a flagship model designed for complex cognitive tasks, scientific reasoning, software engineering, and extended agentic workflows. The new model utilizes greater compute to explore potential solutions and verify intermediate results compared to Gemini 4 Argon, and is being rolled out with introductory API pricing through the end of November.
Adobe adds task agents and Model Context Protocol integration
During Adobe Max, Adobe announced a new wave of AI-powered features across its creative software, including editing and workflow task agents and Model Context Protocol integration. The move enables AI applications to interact in a standardized way with approved creative tools.
Nature features AI agents operating a robotic laboratory
Nature published an original report on a system developed by a team at the University of Toronto that allows AI agents to set up a virtual laboratory and plan robotic procedures before directing actual equipment. The system executed more than 200 commands for a robotic arm to examine the optical properties of nanocrystals.
MWC Doha kicks off with an intelligent connectivity agenda
The second edition of Mobile World Congress Doha kicked off at the Doha Exhibition and Convention Center, co-organized by Qatar’s Ministry of Communications and Information Technology and the GSMA. The opening featured a Qatari agenda to prepare telecom networks for AI agents and the release of two regional reports on the mobile economy and 5G.
AI safety office reports to the White House
An Associated Press investigation revealed that staff at the federal office responsible for evaluating advanced AI safety were recently asked to provide informal reports to a White House office. This arrangement is sparking widespread debate over the independence of government testing and the transparency of oversight for AI agents.