Dogs, dollars, agents

Warhound raises $200M to genetically engineer dogs for the military and law enforcement, a startup conceived by Anduril's Palmer Luckey and Colossal Biosciences co-founder Ben Lamm. The dogs will carry body armor and Anduril's military technology, in a plan that has now come out of stealth with funding.

Why it matters Defense tech money is now flowing into biotech, extending military procurement from hardware and software into engineered animals.

OpenAI nears $30B round at a $1.25 trillion valuation, led by UAE investors, as it finalizes talks. The round is meant as a bridge to a 2027 IPO and would make OpenAI the most highly valued private company ever, in record-setting fashion.

Why it matters A raise at this scale shows how much private capital the leading AI labs still need before they can reach public markets.

Chinese hackers used AI agents to breach ConnectWise, Microsoft says, blaming the China-aligned group Storm-0940 and its ARTEX framework. Microsoft counts 2,250 cloud tenants exposed, and says attackers gained privileged access in some environments in under three hours, using AI agents to do the work.

Why it matters Automated tooling shrinks the time defenders have to react, and a single supply chain breach can reach thousands of downstream customers.

Arc browser shuts down on December 22, as The Browser Company puts its weight behind the AI-focused Dia browser. Arc had a devoted following of enthusiasts but never became the mass-market Chrome alternative its makers wanted, and users are being nudged toward Dia.

Capcom confirms breach of its CAPCOM ID membership service, exposing more than 11.7 million email addresses and associated information. Hackers also changed the login passwords on 196 accounts, so Capcom players may want to be wary of emails claiming to be Capcom.

1Password lets agents log in to your accounts without ever seeing the passwords, with a new product called 1Password for Agents. You still approve sign-ins, and the platform is built for personal agents from Meta, Anthropic, OpenAI and Elon Musk's SpaceXSI.

Why it matters Handing credentials to AI agents is a security risk, and a way to keep passwords out of their reach removes one obstacle to agents acting for users.

China clears DeepSeek, MiniMax to deploy their AI models overseas, a Chinese state media affiliate says. It is part of a wave of homegrown AI technologies that Beijing will allow to operate globally.(paywalled)

Why it matters Beijing is now deciding which of its AI models may be sold abroad, tying China's AI exports to government approval.

Microsoft claims its AI diagnoses better than doctors, with its AI diagnostic system beating generalist doctors on hard cases in a study published in Science. Most of the improvement seems to come from the underlying AI models, not Microsoft's own system, as Mustafa Suleyman pursues medical superintelligence.

Why it matters If the gains come mostly from the base models, similar diagnostic performance may become available well beyond Microsoft.

DeepSeek open-sources another model while China and the US restrict each other's technology, leaving AI leaders wanting their models everywhere and under their control, a pairing they may not get to have.(paywalled)

Canva releases its first image generation model, Canva Craft, which uses a two-stage design that plans a layout before generating the image. It is available to all users and competes directly with models from OpenAI, Microsoft and Google.